Robotics and Computer Integrated Manufacturing 62 (2020) 101861

Contents lists available at ScienceDirect

Robotics and Computer Integrated Manufacturing

journal homepage: www.elsevier.com/locate/rcim

Review

Manufacturing big data ecosystem: A systematic literature review T

a,⁎ a a,b
Yesheng Cui , Sami Kara , Ka C. Chan
a
Sustainable Manufacturing and Life Cycle Engineering Research Group, School of Mechanical and Manufacturing Engineering, The University of New South Wales Sydney,
Sydney, NSW 2052, Australia
b
School of Management and Enterprise, Faculty of Business, Education, Law and Arts, University of Southern Queensland, Springfield, QLD 4305, Australia

ARTICLE INFO ABSTRACT

Keywords: Advanced manufacturing is one of the core national strategies in the US (AMP), Germany (Industry 4.0) and
Smart manufacturing China (Made-in China 2025). The emergence of the concept of Cyber Physical System (CPS) and big data im-
Big data peratively enable manufacturing to become smarter and more competitive among nations. Many researchers
Cloud computing have proposed new solutions with big data enabling tools for manufacturing applications in three directions:
Cloud manufacturing
product, production and business. Big data has been a fast-changing research area with many new opportunities
Internet of things
for applications in manufacturing. This paper presents a systematic literature review of the state-of-the-art of big
NoSQL
data in manufacturing. Six key drivers of big data applications in manufacturing have been identified. The key
drivers are system integration, data, prediction, sustainability, resource sharing and hardware. Based on the
requirements of manufacturing, nine essential components of big data ecosystem are captured. They are data
ingestion, storage, computing, analytics, visualization, management, workflow, infrastructure and security.
Several research domains are identified that are driven by available capabilities of big data ecosystem. Five
future directions of big data applications in manufacturing are presented from modelling and simulation to real-
time big data analytics and cybersecurity.

1. Introduction Customer Relationship Management (CRM) [13] etc. However, smart

Smart manufacturing is critical to national economies by providing
jobs, improving innovation and advancing sustainability [1]. Several
national strategies were initiated to boost their competitiveness of
manufacturing, such as: ‘Industry 4.0’ in Germany [2], ‘Advanced
Manufacturing Partnership (AMP)’ program in the United States [3],
‘Made in China 2025’ and so on. These initiatives provide massive po-
tential to envision the future of manufacturing: Smart manufacturing,
which is defined by National Institute of Standards and Technology
(NIST) as a completely integrated, collaborative manufacturing system
that respond in real time to meet changing demands and conditions in
the factory, in the supply network and in customer needs [4].
Manufacturing industry uses a wide range of software and auto-
mation systems to increase efficiency and productivity from shop floors
to enterprise layers such as CNC machines, Programmable Logic
Controllers (PLC), Supervisory Control And Data Acquisition System
(SCADA) [5], Manufacturing Executive System (MES) [6], product de-
sign and development (CAx: CAD, CAPP, CAM, CAE [7]), Product
Lifecycle Management (PLM) [8], Enterprise Resource Planning system
(ERP) [9], Operating and Maintenance (O&M) [10], Energy Manage-
ment System (EMS) [11], Supply Chain Management (SCM) [12],
manufacturing cannot be realized with traditional manufacturing soft-
ware and technologies due to two main challenges. First, these systems
and software cannot be fully integrated and collaborative since they are
developed by multiple vendors using different interfaces or protocols.
Second, manufacturers cannot perceive and respond to the real-time
changes on time from the factory, supply chain and market since the
traditional manufacturing software lack sensory data to notice the
changes inside and outside the systems.
Digital thread and digital twin are two recent concepts proposed by
integrating disparate systems over the product lifecycle [14] and
building up the real-time relationship between the physical space and
the cyberspace in manufacturing [15] respectively. Cyber-Physical
systems (CPS) are physical and engineered systems, which are mon-
itored, controlled, coordinated and integrated with computing and
communicating core [16]. Internet of Things (IoT) is data-accessing and
data-processing technologies on the cyberspace to perceive the real-
time changes of physical space with sensory tools [17,18]. Digital
thread and digital twin can be enabled by using IoT and CPS. With the
practices of these new concepts and technologies, massive data will be
generated from the systems, which have to go through the processes of
data collection, storage, aggregation, analysis and exchange to provide

⁎
Corresponding author.
E-mail address: yesheng.cui@unsw.edu.au (Y. Cui).

https://doi.org/10.1016/j.rcim.2019.101861
Received 30 January 2019; Received in revised form 28 August 2019; Accepted 10 September 2019
0736-5845/ © 2019 Elsevier Ltd. All rights reserved.
Y. Cui, et al.
Nomenclature
BDA Big data analytic
BI Business intelligence
CAD Computer-aided design
CAPP Computer-aided process planning
CAM Computer-aided manufacturing
CAE Computer-aided engineering
CNC Computer numerical control
CRM Customer relationship management
DSS Decision support system
EOL End-of-life
ERP Enterprise resource planning
HTML Hypertext markup language
IIoT Industry internet of things
JSON JavaScript object notation
KM Knowledge management
MES Manufacturing execution system
MOM Manufacturing operations management
timely information to manufacturers. Being empowered with cloud
computing [19], data science [20] and Artificial Intelligence (AI) [21],
big data focuses on addressing the big data issues amongst the pro-
cesses, which traditional manufacturing tools cannot. Big data could be
the enabler of the concepts of digital thread and digital twin.
Smart manufacturing gains actionable knowledge in real-time with
the fusion of big data and manufacturing knowledge. As big data are
collected and analyzed to extract timely information, manufacturing
industry may still not know which approach to use, and their impacts
without the domain knowledge [22]. The actionable knowledge is
created when manufacturers get timely information from big data and
apply manufacturing knowledge in a specific application. Some ex-
amples found in the reviewed literature are: identifying the reasons of
faults from the production process by analyzing real-time big proces-
sing data and manufacturing knowledge [23], predicting maintenance
intervals by utilizing knowledge discovery and hundreds of machine
data attributes [24], making real-time scheduling and cost-effective
decisions in MES system with streaming shop floor data and existing
manufacturing systems [25]
Similar to big data, manufacturing industry faces the same chal-
lenges associated with 5Vs of big data (Volume, Velocity, Variety,
Veracity and Value) [26]. IDC reports that manufacturing has the lar-
gest share of data (3584 Exabyte) in 2018 and will have 30% annual
growth rate of data from 2018 to 2025 [27]. Of the reported data types,
structured data, such as tabular data in relational databases or
spreadsheets, accounts for only 5% of all the data generated [23]; while
the rest is made up of semi-structured and unstructured data with for-
mats JSON, XML, image, video, and audio, etc. Issues of velocity,
variety and veracity can be explained that the same type of data come
from different devices with various sampling frequencies, formats,
precisions, which leads to inconsistent data and makes challenging to
extract the value-added insight to manufacturers. Amir et al. illustrate
that the limitations of the traditional methods (relational database
management systems (RDBMS) and on-premise software) could not
handle big data [28]. As more manufacturing enterprises generate big
data, the issues of big data will become pressing.
Tapping into the capabilities of big data tools presents enormous
opportunities for smart manufacturing. A large number of big data tools
are developed by the big players in the Internet industry such as
Google, Yahoo, Facebook for their own applications in search engines,
social media, and business analytics [29,30,31], such as Apache Ha-
doop [30], Apache Spark, Apache Flume, Apache Flink, Apache Storm,
NoSQL and NewSQL databases [32], Apache Hive, Apache Pig, Apache
Zookeeper etc. As these tools are enterprise-ready to use in
2
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
MES Manufacturing executions system
NoSQL Not only structured query language
NIST National institute of standards and technology, USA
OLAP On-line analytic processing
OLTP On-line transaction processing
O&M Operation and maintenance
OPC-UA OPC unified architecture
OWL Web Ontology language
PDM Product data management
PLM Product lifecycle management
QMS Quality management system
RDF Resource description framework
RDBMS Relational database management system
SCADA Supervisory control and data acquisition
SCM Supply chain management
STEP Standard for exchange of product data
STEP-NC STEP for numerical control
XML Extensible Markup Language
manufacturing, many researchers reported big data based solutions,
which use a set of big data tools to address problems to enable smart
manufacturing. In 2012, a Hadoop-based sensor data management
framework was proposed for cloud manufacturing [33]. In 2014, Tao
proposed the architecture of cloud manufacturing system as well as the
investigation of applying cloud computing and IoT technology in
manufacturing [34]. In 2015, the Cloud-Based Design Manufacturing
paradigm (CBDM) was proposed by comparing other design methods
[35]. In 2017, Nagorny etc. conclude that big manufacturing data and
big data analytics would provide a vast potential in smart manu-
facturing [26]. In 2017, Wu presented a fog computing-based frame-
work to monitor machine health in cyber-manufacturing. In 2018, Tao
proposed a data-driven conceptual framework to interoperate with
ERP, MES, CRM, PLM systems in manufacturing [36]. Big data tools
complement and provide additional functionalities to address manu-
facturing big data issues which could not have been solved by tradi-
tional approaches.
Three challenging issues have to be addressed in the research of big
data in manufacturing. Firstly, big data tools from Internet industry do
not consider the differences between Internet and manufacturing. Most
manufacturing data is standardized which is supported by various in-
dustrial vendors and associations such as manufacturers of CNC ma-
chines, meters and sensors, controllers and software companies. The
manufacturers use different hardware interfaces, communication pro-
tocols, manufacturing machine readable languages, and semantical
definitions. Whereas, most data in the internet is based on natural
languages and easier to be exchanged without the difficulties associated
with multiple interfaces and protocols. The differences between the two
industries have not been taken into account in developing big data
tools. Secondly, big data tools are massive, diverse, with many overlap
functions. It is challenging to design big data based solution by selecting
suitable big data tools. However, designing big data solutions not only
depends on big data tools but are closely associated with the specific
manufacturing applications and scenarios. This paper categorizes si-
milar types of big data tools and identifies the differences as the pre-
paration to achieve this purpose. Thirdly, many manufacturing systems
have dedicated aims, complicated and sophisticated functions, and are
closely specialized with application scenarios. Social media uses big
data tools to collect and store time series data from billions of customers
who follow each other online; and to report trending events. For
manufacturing, time series data from multiple data sources can be
collected, integrated, and analzsed to explain the states of manu-
facturing entities. For example, the sensor data collected from a CNC
machine reflects the state of the machine and can be used to develop
Y. Cui, et al.
simulation models or prediction models for preventive maintenance. As
many solutions are proposed, a systematic literature review is required
to identify the data issues in manufacturing, capabilities of big data
tools, essential components to design big data based solutions in man-
ufacturing and the potential research directions of big data in manu-
facturing.
The rest of this paper is presented as follows: Section 2 presents the
methodology to systematically review the state of the art of big data
research in manufacturing; Section 3 presents the outcomes of the
systematic review; Section 4 discusses several critical issues of big data
ecosystem in manufacturing including critical drivers, system require-
ments, essential components, research innovation and future directions;
and finally, Section 5 presents the conclusion of this systematic review.
2. Methodology
This paper presents a systematic literature review (SLR) on the
current state of research associated with big data technologies in
manufacturing [37]. To apply big data technologies in manufacturing
successfully, it is essential to systematically review the literature of big
data technologies in manufacturing from the following three perspec-
tives: manufacturing data, big data technologies and data applications
in manufacturing.
Firstly, manufacturing data is the foundation to conduct data-driven
manufacturing. It is impossible to propose one big data based solution
to fit all manufacturing circumstances since different applications have
different data issues (data types, data formats and data sources) and
require specific tools to address. Therefore, systematically analyzing
manufacturing data could provide a useful guideline to select appro-
priate big data enabling technologies.
Secondly, the big data tools in the big data ecosystem have to be
identified the similarities and differences. The 5Vs characters of big
data are widely recognized as challenges, such as volume (TB/PB level
of data size), velocity (ingesting or processing big data in streams or
batches, in real time or non-real time), variety (dealing with complex
big data formats, schemas, semantic models and information), value
(analyzing data to deliver added-value to some events), and veracity
(validate data consistency and trustworthy) [38]. In general, these big
data technologies are intended to address some Vs of big data. Hence,
Fig. 1. Conceptual framework of systematic literature review.
3
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
their capabilities need to be effectively classified and analyzed to know
which Vs are addressed.
Thirdly, gaps of data applications in manufacturing could be iden-
tified by systematically reviewing the capabilities of the traditional
manufacturing systems and big data analysis. Since much traditional
manufacturing software has been widely used in enterprises, big data
could integrate and collaborate the software and systems as well as
providing timely information. The massive amount of data generated by
these applications can be fed back to the big data ecosystems for ana-
lytics and innovative applications such as prediction, optimization,
monitoring, simulation and visualization, etc. Therefore, these appli-
cation gaps would be the future research directions in academia and the
demands in the industry.
In summary, knowing the data requirements of manufacturing ap-
plications, understanding the capabilities of big data tools, and identi-
fying the gaps will help define future research directions and generate
new ideas for innovative applications. This systematic literature review
presents a holistic overview of big data in manufacturing to study the
possible use cases for manufacturing. As shown in Fig. 1, the conceptual
framework of this systematic literature review includes three layers:
data source, big data ecosystem, and data consumers.
The first layer, at the bottom of Fig. 1, is the data source, which
consists of five aspects:
1 Data types refer to the meaning of data such as temperature, hu-
midity from the physical space, and log, email, operational data
from cyberspace;
2 Source devices to collect data sources, which include sensors, con-
trollers, actuators, software systems. The data type and source de-
vices have a close relationship with the first four characteristics of
big data (Volume, Velocity, Variety, Veracity) [39]. For example, in
order to know the temperature of a production line, a temperature
sensor is selected with the determined sampling rate (speeds of data
generation), the sizes of data accumulating by time, formats and
quality of data from the sensor.
3 Data dynamics describe the states of data. Data-at-rest refers to the
inactive data stored in spreadsheets, databases and data ware-
houses; while data-in-motion refers to the active data generated by
sensors, equipment or machines, and fed into the big data ecosystem
Y. Cui, et al. Robotics and Computer Integrated Manufacturing 62 (2020) 101861

Fig. 2. Process of literature review methodology.

Table 1
Data formats in reviewed articles and NIST report.
Categories Systems Structured Semi-structured Unstructured

Product CAD/CAE/CAPP/ [53] XML [54], G-code [55], STL [56], (aml, obj, UML, AutomationML) [57], IGES, DXF, AMF, RDL ○
CAM
PLM [58] (XML, B2MML) [59], PMML [60], (RDF, SPARQL, STEP, QIF, STL) [61], PLMXML [60] pdf [61]
Production ERP [62] (XML, HTML, SCUFL) [63], PMML e-mail [64]
MOM/MES [65, 66] JSON [67], RDF [68], AutomationML [69], BatchML ○
SCADA/DCS/HMI [70] (RDF, OWL, XML) [71], SPDML [72] Image [73]
IIoT/CNC/Robot ○ (XML, UML, AutomationML) [69], OPC-UA, PLC Open, COLLADA) [74], BSON [75], JSON image [77, 78],
[76]
O&M [79, 80] cad [80], RDF [81], XML [82] Image [83], video [82]
QMS [84] BSON [85], (XML, QIF) [86] image [87], (audio, document)
[88]
Safety [89] ○ ○
Business SCM//BI/AM [90] (XML, JSON, RDF) [91], (WSDL, EPL) [92], XPDL, ebXML, BPEL, UBL, WS-CDL, OAGIS, document [93]
ICT CPS/CM/ICT [94,95] (JSON, STEP, JT Open) [96], HTML [97], (AutomationML, PLCopen) [98], (XML, XSD, RDF) e-mail [64]
[99], (UML, SysML, STEP, B2MML) [5], EDDL [100,101]
Data analytics/DM [102] BSON [102], JSON [103], Parquet [104] (image, video, document)
[105]
KM ○ (OWL, UML, RDF, SWRL) [23], (JSON, PMML, AMPL) [106], (STEP-NC, G-code, XML, DMIS, ○
QIF) [107]

○: Not found.
Text: Not found in reviewed articles (mentioned in the NIST report).

in real-time. • Computing engines: batch processing (MapReduce), iterative/near

4 Data formats are structures of the data. Data is exchangeable among
various systems with consistent data formats and languages.
The system aspect of data source refers to the system where data is
originated. There is a diverse range of manufacturing systems used in
different applications such as product design, manufacturing pyramid,
product lifecycle management, supply chain management, logistics.
The second layer, the middle layer shown in Fig. 1, is the big data
ecosystem comprising all the big data software. This layer plays the role
of connecting the data sources from the layer below and the big data
analytics applications at the layer above.
The big data ecosystem is a set of complex and interrelated com-
ponents to process and analzse big data [40]. Also, the ecosystem needs
to store data from various data sources for data integration and ana-
lytics as well as other applications. Therefore, data storage layer in the
ecosystem includes database and file system technologies to store big
data. The ecosystem consists of the following components and tools:
• Data collection and ingestion: log data collection (Flume), bulk data
collection from a relational database (Sqoop), distributed messaging
system (Kafka), dataflow (NiFi);
real-time processing (Spark, Flink), real-time processing/streaming
(Storm, Flink) [41];
• Database: Relational database (RDBMS) has a standard schema but
without scalable capabilities (MySQL, Oracle DB, SQL server,
ProgresSQL); NoSQL database does not have a standard schema and
has scalable capability (four types NoSQL: Column-based: HBase;
Document-based: MongoDB; Key-value-based: Redis; Graph-based:
Neo4j); NewSQL database is scalable relational database (VoltDB)
[42,43]), search engine (Solr, Elasticsearch);
• Data analysis (BDA): Machine Learning (MLlib, Caffe, Tensorflow,
Python), statistic (SparkR, R), OLAP,
• Data visualization (Zeppelin, Matplotlib, Tableau, D3 [44], GraphX;
• Workflow which is a scheduler of the jobs of various big data tools
and dataflow which manages data transfer and data transformation
among different big data tools: Oozie, Kepler, Apache NiFi;
• Data management and KM: Apache Falcon, Apache Atlas, Apache
Sentry, Apache Hive, Operation (Zookeeper, Ambari), Apache
Griffin, Apache Ranger, Apache Jena;
• Big data infrastructure (BDI): computing resources (general purpose
computing and HPC), cluster management (YARN, Mesos) [37],
network communication (Software-Defined

4
Y. Cui, et al. Robotics and Computer Integrated Manufacturing 62 (2020) 101861

Fig. 3. Smart manufacturing systems and various data formats.

Table 2
Data issues in manufacturing.
Issues 5Vs Description References

Large scale Volume, Variety Large volume dataset with massive features. [109]
Inconsistent sampling frequencies or Velocity Sensors use various sampling frequency and timestamps; [110,111,112]
timestamp Unnecessary high sampling frequency affect the real time performance, it is related with [113]
“Smart data” topic.
Batching data and streaming data Velocity Data modification is sensitive to time or not, some literature also uses the terms: data-at-rest [114]
and data-in-motion [13];
Missing value Veracity Record is empty when equipment is an anomaly; [115]
Product passes some portion of machines; [109]
Sensor is an anomaly or losing communication with sensor; [110]
Too costly to capture data by installing sensor or building models. [113]
Imbalance Veracity Small probability data in a very large dataset with most of normal data. [109]
Data outlier Veracity Data is out of range of measurement device. [110,84]
Noise or an anomaly data Veracity Data is out of the similar clustering dataset; [115]
Noise data is possibly generated by replacing the missing data. [109]
Drifting data Veracity Process drift caused by vulnerability to external environment; Sensor drift caused by [110]
modification in measuring device or calibration.
[113]
Asynchronization Veracity Several data producers (sensor and machine) use non-central time server in manufacturing [113]
enterprise.
Data correlation Veracity Nature and structure of data caused by redundant sensor arrangement; [110]
Data correlation is to improve data quality with process variables. [116]
Inconsistent simulation and collection of data from CNC machine. [107]
Data model, data format exchange and Variety, Veracity, Merge data from multiple data sources into a single view; [115]
data integration Value Exchange different data formats from various systems: OPC-UA and IoT, OPC-UA and [117,98,86,[118],
AutomationML, MTConnect with QIF, MTConnect and IEEE 1451 wired smart transducer
Integrate data models to explain the relationships of data; [113]
Data information must be available for information sharing. OWL is the enabling [99,118]
technology on resource description.

• Networks (SDN) [45], InfiniBand [46], 5G [47]) etc.;
• Big data security: Apache Metron [44], Apache Knox;
Although Hadoop is a big part of the big data ecosystem with many
big data tools [48], it lacks functions such as data flow, data manage-
ment and security [41].
Finally, the top layer represents the way data is used and data users.
It includes the applications of big data analytics [26] and manu-
facturing applications [36]. The databases of traditional manufacturing
software maybe the data sources of the ecosystem as well, such as
SCADA or EPR.
2.1. Literature identification
Pertinent articles are identified within the scope of manufacturing
data, big data technologies and big data based solution in manu-
facturing. Fig. 2 illustrates the research method in this literature re-
view. First, four citation databases are chosen due to their compre-
hensive coverage and high relevance to the scope: Scopus, IEEE Xplore,
ASME Digital Collection, and ACM Digital Library. Second, several big
data technologies and popular manufacturing data collection tools are
selected as keywords. Hadoop is chosen since it is the earliest big data
technology which is well studied and used. The underlying technologies

5
Y. Cui, et al. Robotics and Computer Integrated Manufacturing 62 (2020) 101861

Fig. 4. Chronological distribution of big data tools.

Fig. 5. Percentage allocation of Big Data applications (Based on our literature investigation).

of big data are computing and storage. There are a few big data com-
puting engines. Three common ones are selected: Spark, Storm and
Flink. For storage, because there are hundreds of available databases, it
is not appropriate to limit to specific databases. All the databases could
be categorized into three types: SQL, NoSQL and NewSQL. We select
SQL and NoSQL because NewSQL could be recognized as SQL database
with better features than traditional SQL. The features of NewSQL will
be discussed in the following Sections. Time series data widely exist in
manufacturing such as machines, sensors, controllers. Time-Series da-
tabase (OpenTSDB) is selected since it seems more suitable for manu-
facturing data. Two widely adopted manufacturing data collection tools
are selected: OPC-UA [49] and MTConnect [14].

6
Y. Cui, et al. Robotics and Computer Integrated Manufacturing 62 (2020) 101861

Third, to further focus on the nature of this research paper, the

papers were filtered by using the abstract “Manufacturing”, “Industry

logistic regression, naïve Bayes, and a decision tree [109], regression [137], LSM [138], SVM [83], Anomaly detection [139], DTW
4.0”, “Industrial automation”, “Smart manufacturing”, “Digital twin”
and “Digital thread”. Four, manual review is implemented to select the
papers about manufacturing data issues, or big data based solutions and
applications in manufacturing. For example: some papers of other in-
dustrial sectors are found since they merely mentioned manufacturing
in the abstract, such as Oil and Gas, Healthcare, Energy and
Agriculture. Hence, the 339 articles are reviewed and 128 relevant
articles are selected. The search strings of four databases are listed in
Table 1 in Appendices.
regression [127], Distance, Regression, Self-organizing map, principal component analysis [128],

3. Results

3.1. Manufacturing systems

In 2016, NIST reported three dimensions of concerns in smart

manufacturing systems (SMS): product, production and business. Many
traditional manufacturing systems and software can be categorized into
one of the dimensions [49] (Fig. 3). Business dimension is presented in
the upper rectangle block with dash lines, which includes suppliers,
SPC [87], ANN, decision tree, random forest, SVM [143, 84]

customers and manufacturing enterprises (SCM, CRM, BI, asset man-

agement). Product dimension is located at the bottom rectangle block
Classification [133], OPL [106], KM [134], GA [73],

with solid lines. It includes objects and activities from product design to
[140], RF [141], K-means, Markov [142], KD [24],

end-of-life of the product (CAx: CAD, CAM, CAPP and CAE, PLM).
KM [106], Semantic data integration [23, 153],

Production dimension is the triangle block, which includes an entire

MP [63], regression [62], K-means [126]

production system (ERP, MOM/MES, SCADA/DCS/HMI, O&M, Safety,

quality management). Industry Internet of Things (IIoT) and RFID
technology (grey blue) are widely used in manufacturing, logistics [50],
in-use and product End-of-Life (EOL) [51].
Regressions [53], ANN [53],

Stream data analytics [103],

Afterwards, all the reviewed articles are classified into four cate-
gories: Product, Production, Business and ICT (Information
graph analytics [89],

Communication Technology). The first three categories focus on en-

Big data analytics

gineering functions and business, ICT architecture underpins all three

dimensions to provide the ICT infrastructure and digitalization to
manufacturing, which includes several topics: CPS, Cloud manu-
facturing (CM), ICT, Data analytics/Data management(DM), KM.
Allocations of proposed solutions of reviewed literatures (Based on our literature investigation).

Table 2 in Appendices illustrates the distribution of reviewed articles by

Big data computing and storage

these four categories.

[104,130,131,104,132,112]

3.2. Data source

[68,148,95,149,150],

3.2.1. Data format

[136,81,82,80]

Based on the three dimensions of SMS in NIST report, the standards

[57,56,125],

of data formats, computer languages are listed with the italic style in
[152,94],
[144],

every manufacturing system in Fig. 3. The bold black text represents

[145]
[66],

some of the data formats found in the reviewed articles, whereas the
black text is not found in the review but mentioned in the NIST report
Big data framework

[49].
[64,146,76,147]

Table 1 demonstrates the complete data formats found from the

reviewed articles and NIST report. To discuss the formats conveniently
[69,116],
[129,71],
[39,135]

[151,33]
[93,90],

in the following chapter, all the data formats are categorized into three
[58],

groups:

• Structured data: data that is presented in tables and can be stored in

Data analytics/DM
SCM/CRM/BI/AM
IIoT/CPS/CM/ICT
SCADA/DCS/HMI
CAD/CAE/CAPP/

a relational database;
• Semi-structured data: data that has a self-described structure and is
MOM/MES

not presented in tables, such as XML, JSON, HTML [52];

• Unstructured data: data that does not have a self-described struc-
Systems

Safety
O&M
CAM

QMS
PLM
ERP

KM

ture, such as document, image, audio, video, text and e-mail.

ICT architecture

Fig. 3 and Table 1show that one challenging issue to realize that
smart manufacturing requires different data formats from various
Production

Business

manufacturing systems. In order to make these systems collaborative

Product
Table 3

and integrated, the transformation of these data formats is an essential

Field

function to the manufacturing big data solutions. It also illustrates that

7
 Y. Cui, et al.

Table 4
Summary of discussion section.
Topics Sub-topics Application systems Enabling tools References

6 drivers for big data in smart
manufacturing
9 essential components of big
data ecosystem
System integration Product design, AM, ERP, MES, BI, SCM, PLM, Kepler, Hadoop, OPC-UA, RESTful API [49,57,56,125,63,62,90,153,97,55,128,154,155]
Data Predictive maintenance, KM, Production planning, Cassandra, MongoDB, Blueflood, [121,62,73,156,157,82,138,158,24,89,94,153,159,75,160,88]
Safety, Anomaly detection, industrial process control, OpenTSDB, DalmatinerDB and InfluxDB,
model prediction, QC, shop floor scheduling, Storm, Spark, Flink, Apache Hive,
Prediction 3D printing, product performance, production Random Forest, Bayesian Network, statistic [39,109,63,62,93,136,137,138,141,24,87,143,90,55,161,85,88,25,162,79]
planning, energy consumption, MES, QC, SCM
Sustainability PLM, Maintenance [163,59]
Resource sharing and SCM, ERP, Data integration Public cloud, Private cloud, Hybrid cloud, [19,63,62,144,95,152,97,101,164]
networking hypervisor, container,
Low cost hardware SCM, Industrial automation RFID, IoT, Robotic [65,165,166,167]
Data ingestion ERP, MES, SCADA, O&M, QC, PLM, Data management Sqoop, Flume, Kafka, [39,116,58,71,138,139,140,151,85,79,65]
Storage ERP, SCM, SCADA, OLAP, OLTP Redis, HBase, Cassandra, MongoDB, Neo4j, [30,32,43,168,169]
HDFS, VoltDB, Clustrix, NuoDB
Computation ERP, SCM, PLM, MES, SCADA, O&M, QC, IoT MapReduce, Spark, Flink, Storm, [170,171]

8
Analytics DSS, CRM, machine vision, QC, O&M MLlib, Scikit-Learn, CNTK, Caffe, Kylin, [109,73,82,87,25,105,172,173]
CaffeOnSpark, Hive, CaffeOnSpark
Visualization MES, SCADA, O&M, QC, Security, Zeppelin, Tableau, D3.Js, Matplotlib, [126,76]
QlikView
Workflow and dataflow Business Oozie, Kepler, InfoSphere, Wings/Pegasus, [116,123,63,172]
NiFi
Data management ICT, KM, SCADA, SCM, Apache Falcon, Apache Atlas, Apache [121,23,174,175,176,177,178,179]
Sentry, Apache Griffin, Jena
Infrastructure and ICT HPC(AWS), [180,181]
deployment model
Cybersecurity SCADA, ICT [182,183]
5 future directions Modelling and VR/AR, PHM, PLM, CAx Cloud computing, Quantum computing [184,185]
simulation
Connectivity and File formats of PLM (PLM XML) and 3D printing (AMF, NiFi, [186]
interoperability 3MF)
Standardized big data [5]
platform design
Real time big data SCADA, MES, Data warehousing Spark, Storm, Flink, Beam, Spark R, MLlib, [26,187]
analytics GraphX, SparkSQL
Cybersecurity SCM, SCADA, Security, Safety Apache Metron, Apache Ranger, Apache [49,188]
Knox
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
Y. Cui, et al.
some data formats in the specific manufacturing systems are missing in
the proposed bid data solutions. It requires solutions to fill the gap to
realize data exchange among the systems. Big data tools can address the
variety issue of manufacturing data such as NiFi.
3.2.2. Data issues
Data issues are fundamental challenges to smart manufacturing,
which extract actionable information from good quality of data. In
order to prepare the suitable data for smart applications, amount of cost
and time is consumed to address the data issues. For example, data
scientists spend over 90% of their time on data preparation before
analyzing data for innovative tasks such as machine learning, AI [108].
Enterprises spend billions of dollars on their data warehousing systems,
which can only use well and pre-defined methods (ETL) to process
product structure data and produce business reports in non-real time.
Therefore, understanding and addressing data issues is critical to design
big data-based solutions. From Table 2, we have identified 11 common
issues of manufacturing data. Data issues are generally related to the
Big data 5Vs features [38]: volumes and variety (large scale); velocity
(inconsistent sampling frequencies or timestamp, batching and
streaming data); veracity (missing value, imbalance, data outlier, noisy,
drifting, asynchronization, data correlation); variety, veracity and value
(data model, data format exchange and data integration). The re-
searchers of these kind of literature address these issues without using
big data tools. Traditional software cannot address these data issues if
the data is generated in large scale systems with large number of de-
vices. Some big data tools are discussed to provide potential solutions in
Sections 4.2.7 and 4.4.2.
3.3. Big data ecosystem
Through this SLR, new big-data based research innovation could be
identified by closely tracking the attention of various big data tools in
manufacturing and other research domains. Fig. 4 shows the distribu-
tions of big data tools over the years from the literature, including
manufacturing engineering and others. Fig. 4(a) shows the numbers of
the term Hadoop occurred in all literature increased from 2008,
reached its peak in 2016 and slightly decreased in 2017. The number of
Spark articles had been higher than Hadoop since 2016. There are four
factors to compare Spark and Hadoop: volume, velocity, fault-tolerant
and data analysis. MapReduce and Spark Streaming are the computa-
tion engines of Hadoop and Spark, respectively. MapReduce executes
batch processing by reading and writing data on disk multiple times.
Spark Streaming executes micro-batch processing in memory. It results
in the differences between Hadoop and Spark because disk can persis-
tently store a larger volume of data with slower velocity than memory,
which temporarily stores limited volume of data. By comparing data
analysis, Spark has built-in tools (MLlib) and support third-party tools
(Mahout, H2O), whereas Hadoop is only supported by the third-party
tool: Mahout [41]. Spark supports iterative computation with GraphX,
which is the graph processing engine. Therefore, Hadoop is suitable for
the applications which need planned extraction of non-real time and
critical information from a larger volume of data and guarantee without
loss of data, such as ERP, Production planning. Spark can be used to
provide near real-time monitoring and analytics by processing
streaming data such as monitoring process and product quality, MES,
SCADA, predictive maintenance. The number of Storm articles has kept
increasing since 2012; however, its total number is smaller than Ha-
doop and Spark since it only executes streaming with limited data
analysis supported by SAMOA, which is a version of Mahout [119].
Fig. 4(b) presents the growing number of literatures about these three
tools in the manufacturing research literature since their first inception.
From Fig. 4(c) and (d), NoSQL database, as the highest frequently
mentioned database, is compared with NewSQL and time series data-
bases (OpenTSDB) over the last eight years. Fig. 4(e) also demonstrates
that Kafka and OPC-UA are closely related to manufacturing as seen by
9
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
their increasing patterns. Because OPC-UA protocol provides open
connection to monitoring and automation systems as well as commu-
nications between MES and SCADA systems, these systems are pre-
valent in manufacturing [120]. Kafka is also used widely in applications
at the shop floor level of manufacturing such as processing, machine
and sensor data, due to it is message streaming capability [121].
Apache NiFi is entirely new to manufacturing. Only four articles were
found in the industry in general and no article in manufacturing in-
dustry [122,123,124].
3.4. Applications of big data in manufacturing
Data applications are essential to realize Smart manufacturing.
Identified scopes of data applications could provide a clear guideline to
design manufacturing big data platforms. Among the reviewed 128
articles, 78 articles are big-data based applications, which are cate-
gorized into17 applications. The percentages of the applications are
presented in Fig. 5. Monitoring (25%), prediction (23.8%), ICT frame-
work (11.9%) and data analytics (9.5%) are the four most frequently
used big-data applications in manufacturing. Through this statistic, it
can be shown that the researches of big-data based solutions focus on
monitoring, prediction, data analytics and propose ICT solutions in
manufacturing in Table 3.
4. Discussion
Four fundamental questions about the relationship between big data
ecosystem and smart manufacturing are six drivers and requirements
for big data application in manufacturing, seven essential components
of the big data ecosystem, harnessing big data capabilities for research
innovation in manufacturing, and future directions of big data appli-
cation in manufacturing. They illustrate the driving factors of big data
applications in manufacturing. These are summarized in Table 3 and
Table 4 for ease of reading to interested readers.
4.1. Question 1: what are the drivers and requirements for big data
applications in smart manufacturing?
Identifying drivers for big data applications is essential to imple-
ment feasible smart manufacturing initiatives. Kusiak discusses the fu-
ture developments in manufacturing and identified six drivers for smart
manufacturing theoretically [4], which are manufacturing technology
and processes, material, data, predictive engineering, sustainability,
resource sharing and networking. Through reviewing the proposed big-
data based solutions, six drivers are identified: data, prediction, sus-
tainability, resource sharing, system integration, and low-cost hard-
ware. Four common drivers from Kusiak's six drivers are verified with
big-data based solutions.
4.1.1. Driver 1: system integration
System integration with big data technologies is a crucial enabler of
smart manufacturing to integrate and cooperate manufacturing systems
to timely adapt dynamic demands from production and supply chain
[49]. Integration of production systems demonstrates the significant
improvement in production efficiency and productivity since the 1980s
[4]. In the context of smart manufacturing, it needs to further expand
the scope of system integration from production to product and busi-
ness domains. Fig. 3 shows that various manufacturing systems use
different networks and protocols, which are challenging due to the
necessity of implementing data and information exchange among these
systems. Big data technologies can integrate these independent systems
by using “cloud” as a common place to collect data, extract and ex-
change the required information on the cloud. With the fusion of IoT,
BDA not only can integrate manufacturing systems but integrating
physical and cyber worlds closer [36]. From the system integration in
Table 4, many benefits of system integration are demonstrated with the
Y. Cui, et al.
proposed big-data based solutions.
Integrating product design and additive manufacturing with big
data provides many benefits. With the integration of product design and
additive manufacturing, product costs can be estimated by analyzing
more features of product model with DBA than the traditional approach
[55]; using Hadoop clusters demonstrates faster velocity of converting a
huge 3D model to G-code for 3D printer than traditional methods [56].
Spark and Cassandra demonstrate capabilities of computing and storing
a large volume of streaming data from the application of real time
monitoring 3D printing, which traditional manufacturing systems
cannot offer [125].
As to production, big-data based solutions drive the integration of
ERP and MES. The big data scientific workflow management system
(Kepler) demonstrates the efficient scheduling capability for smart
manufacturing [63]. Another paper concludes that the critical cycle
time can be predicted with Hadoop for production planning [62].
In the business domain, several researchers focus on designing big
data architecture to integrate BI, SCM, ERP, MES and PLM systems.
Although a business intelligence architecture is proposed to advance
the integration of business information from various existing systems
such as ERP, CAx, SCM, PDM/PLM, the specific technical framework is
not provided [90]. It demonstrates that it is necessary to have a cap-
ability to store data and information with various data formats, struc-
tures and models to integrate various systems. Another big-data based
solution is proposed to integrate supply chain and production planning
by retrieving and integrating SCM, ERP and MES data [97]. However,
the framework focuses on designing business functionality without
providing support to process the collected big data. A cloud manu-
facturing collaboration system is proposed to achieve better perfor-
mance and functionality on production, resource planning with the
Hadoop ecosystem. It is concluded that information integration from
the web and other data sources is a critical issue to implement system
collaboration [153]. Processing and analysing data from MES and
SCADA with MapReduce and BDA can detect anomaly minutes be-
forehand in a large-scale production [128]. In the industry, Bosch
presented a conceptual, analytic platform with a data integration
method to integrate various data sources [154].
System integration with BDA requires standard data format and
standard interfaces. New standards of file formats need to be developed
in order to fill the deficiencies of the existing standards to transfer
consistent content [49]. Additive Manufacturing File (AMF) is a new
XML based standard format to replace STL by providing many new
features in addictive manufacturing such as materials, material prop-
erties, colors. [49]. Because RDBMS is not suitable to store data with
flexible data models (unstructured or semi-structured data), big data
technologies are thus mainly used to process unstructured data such as
Hadoop, NoSQL databases. The standard interface is of importance to
seamlessly integrate systems in manufacturing. Because OPC-UA tech-
nology provides dedicated interfaces to production equipment such as
PLC, it received increasing attention in the industry [128]. RESTful API
demonstrated to be more efficient to connect web data by comparing
with the traditional method SOAP [155].
4.1.2. Driver 2: data
Timely comprehensive data with enabling big data tools is the key
driver to smart manufacturing. The data row in Table 4 presents many
manufacturing applications could be implemented with more compre-
hensive data and big data tools, which were challenging with tradi-
tional tools. Firstly, a large volume of data collecting from various
sources provides sufficient data for big data analytics. Data from var-
ious industrial equipment, IoT devices, web and smartphones, is called
data-in-motion [159], which is continuously generated and ingested
into the systems to provide real-time response from the physical world,
such as dynamic shop floor scheduling [75], predictive maintenance
[141,136,24], anomaly detection [81], diagnosis [82], prognosis [161]
and systems collaboration [153]. As another significant data source for
10
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
big data ecosystem, databases in manufacturing is data-at-rest, which
represents static, historical data [159]. This data is mainly used to
predict the long-term performance in production planning [62], global
manufacturing network design [94], critical event detection in safety
[89]. Both historical batching data and real-time streaming data are
integrated to train models and monitor real time condition information
such as anomaly detection of machines’ energy consumption data
[139].
Secondly, big data technologies make the management of manu-
facturing big data feasible. Traditionally, RDBMS is mainly designed to
store structured data with limited scalability. However, NoSQL data-
bases present better performance for handling semi-structured (JSON,
XML) and unstructured data (audio, video, and email) with unlimited
scalability. For example, column NoSQL database Cassandra was used
to store event data of automation controller [131], document NoSQL
database MongoDB was used to store machine data [85]. Time-series
databases (TSDB) begin to receive increasing attention by providing
dedicated applications for sensor data. A comprehensive evaluation was
implemented to several TSDBs: Blueflood, OpenTSDB, DalmatinerDB
and InfluxDB [82]. The collected data needs cleaning before usage in
order to resolve issues such as noisy and incorrect format, as shown in
Table 2. Streaming (Flink, Storm), micro-batching (Spark) and batching
(MapReduce) data processing technologies provide the capabilities to
clean and calculate big volume of manufacturing data. The following
big-data based solutions are proposed and implemented in various
manufacturing applications: complex event processing (CEP) with
Storm [138], anomaly detection with Flink [121], industrial process
control with Spark [73], model prediction [137] and quality control
with MapReduce [88].
Lastly, past and new knowledge can emerge from the generated big
data by harvesting big data technologies. Knowledge of predictive
maintenance can be extracted with an Apache Hive-based platform
[24]. Knowledge of intelligent applications of a smart factory is man-
aged with Hadoop and OWL technologies [134].
4.1.3. Driver 3: prediction
Prediction enables manufacturing to change from reaction to pre-
vention. Because of big data and increasing applications of data analysis
in manufacturing, it is feasible to predict the behaviours of various
manufacturing systems accurately.
Prediction attracts many researchers’ attention to manufacturing.
From Fig. 5, the prediction is the common BDA application in manu-
facturing. In the product domain, the costs of 3D printed products can
be predicted with the proposed big-data based solution and three ma-
chine learning algorithms [55]. In the production domain, machine
learning is used for the prediction of product performance degradation
[63], cycle times for production planning [62], energy consumption
and KPI values on MES system [25], and production efficiency [162]. A
number of papers focus on the prediction of product quality with dif-
ferent machine learning algorithms such as Bayesian Network [88] and
statistic analytics [87]. ANN is identified as the highest prediction ac-
curacy by comparing with a decision tree, random forest and support
vector machine [143]. Some applications need a trade-off between
acquiring higher accuracy and shorter calculation time since it takes
more time to calculate to get higher accuracy, such as real-time quality
control. In terms of the full consideration of shorter calculation time
and higher accuracy, Random forest is outperformed Naive Bayesian,
Multi-Layer Perceptron and Logistic Regression [85,109]. Many papers
publish the solutions of predictive maintenance on machines by using
machine learning algorithms to recommend scheduling of proactive
measures before outages occurred [24,39,79,136–138,141,161]. In the
business domain, some potential applications are found such as pre-
dicting user behaviour with using principal component analysis and
Hadoop [128], proactive inventories, location and throughput times on
logistics with a big data based platform [90].
Y. Cui, et al.
4.1.4. Driver 4: sustainability
Sustainability with big data technologies plays a vital role in smart
manufacturing. Sustainable manufacturing considers the four factors:
material, manufacturing processes, energy and pollutants. Big data
technologies can provide a data-driven solution to analzse the big vo-
lume of data for these four factors. For example, product design could
be guided by analyzing the End-of-Life data of products; strategic de-
cision making by analyzing marketing, production and supply network
data from CRM, ERP, MES and SCM systems; energy consumption and
pollutant influence would be monitored through IoT sensors and RFID
tags. Although several conceptual big-data based frameworks are pro-
posed as shown in Table 4 [58,59], the performance of sustainability
with big data technologies has not been evaluated. Further research is
required to discuss the effectiveness of big data technologies on sus-
tainability in manufacturing.
4.1.5. Driver 5: resource sharing and networking
Manufacturing could benefit by sharing virtual and physical re-
sources with the supply chain network. The issue of information silo
results in losses of productivity and economy in manufacturing.
• Sharing information
Sharing information is beneficial to manufacturing systems of pro-
duct, production and business. Helu et al. discuss that digital thread can
improve product design and manufacturing processes by sharing data of
product lifecycle systems [14]. For instance, some information on idle
equipment among enterprises can be shared with big data tools in order
to reduce holding costs such as machines, 3D printers, equipment,
transport and warehouses [101]. Sharing information about SCM and
PLM systems could meet the new business requirements such as find
best supplier of a specific raw material [164]. Integrated data from
various systems could identify potential production problems and im-
prove work efficiency by selecting suitable maintenance time [152].
Collection and sharing data of supply chains with big data could be
helpful to interactions among customers, manufacturers, and suppliers
[97].
Addressing information silo is challenging because data cannot be
easily shared among traditional manufacturing systems by different
protocols. Some practical solutions are implemented, such as sharing
data in various systems, which leads to the issue of data redundancy.
Manufacturers also have to take massive maintenance effort on syn-
chronizing the shared data of every system. Because the big data plat-
form could be sat on the cloud, it has to establish only one connection to
synchronize data of the platform and the system. Through big data
tools, data is much easier collected from various systems to the data
lake of the big data platform, synchronized, managed and shared [121].
• Sharing BDA Infrastructure and software
Sharing big-data infrastructure and software brings economic ben-
efits to manufacturing. According to NIST definition, there are four
deployment models of cloud computing: private cloud, community
cloud, public cloud and hybrid cloud [19]. From reviewed papers, 18
solutions use private cloud while only two solutions use public cloud.
The main reason for adopting a private cloud is to provide better
privacy and security than public cloud [62,63,144]. However, there is a
significant investment on the hardware of private cloud clusters. Wang
et al. demonstrate that manufacturing can gain economic benefits from
the public cloud with three aspects: pay-as-you-go service model, re-
ducing maintenance fee on data centers and operating cost [189]. No
security function was identified in the 18 proposed solutions, which
means the security and privacy of the private cloud solutions do not
outperform public cloud. Therefore, using the public cloud in com-
parison to private cloud may bring significant economic benefits to
manufacturing.
11
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
Virtualization technologies such as hypervisor and container, pro-
vide faster deployments, high efficiency to share software packages in
enterprises [95]. Virtualization can be used to quickly test and validate
the proposed big data solutions with minimum influence on the other
systems.
4.1.6. Driver 6: low cost hardware
Low-cost hardware makes smart manufacturing more accessible.
Low-cost actuator and IoT sensor reduce the wiring cost to collect data
and improve automation at the factory floor. Wang et al. present a
collaboration mechanism to deploy large scale robotics with big data
technologies at a small factory [65]. Big-data based solution with smart
sensors elevates the constraints of time and geolocation to monitor
manufacturing processes [165]. Low cost RFID tags make traceability of
enormous resources more feasible at the supply chain level. Industrial
Internet of Things hub is proposed to realize the smart connection of
various resources in a manufacturing facility with RFID tags and sensors
[166]. Cost competitive NC machines and 3D printers can be flexibly
deployed to demand sides such as design prototype product or serving
customers. Big data technologies can secure critical data transmission
between design departments and manufacturing equipment. Further-
more, low-cost data processing and storage technologies provide a cost-
effective approach to manage large volumes of data from massive data
sources in manufacturing [167]. Hence, cheap hardware makes it more
feasible for manufacturers to monitor and respond to timely changes
from production, to supply chain networks.
4.2. Question 2: what are the essential components of big data ecosystem to
better serve smart manufacturing?
As big data applications enable smart manufacturing, several es-
sential components of the big data ecosystem should be utilized to build
up BDA platform for smart manufacturing, including data ingestion,
storage, computing, analytics, visualization, workflow and dataflow,
data management, infrastructure and security.
4.2.1. Data ingestion
Data ingestion or inception is of necessity to manufacturing in order
to bring big volume data into its BDA platform. There are two types of
big data: data at rest and data in motion. Apache Sqoop is used to
transfer bundle of data from a relational database (MySQL, SQL server)
to Hadoop in several applications such as ERP and MES [116,65],
SCADA [71], O&M [39,79] and Data Management [121,151].
Streaming data is data continuously generated from manufacturing
systems and devices. Apache Flume is mainly used to collect large
amounts of logs from controllers, sensors, equipment and actuators
[65,71,116,121]. Kafka is a general-purpose messaging system to col-
lect streaming data and publish it to data consumers who subscribe to
the topic of data. Kafka is applied in some cases of SCADA [140], O&M
[139]. Apache Storm is a real-time data processor, which is used to
collect and ingest streaming data to data consumers straightway (O&M
[138,139], Quality control [85], PLM [58]). Although some data col-
lection tools of manufacturing are widely used in production systems
such as MTConnect and OPC-UA, big data ingestion tools can comple-
ment their limitations as discussed in Section 4.4.2. Therefore, data
ingestion is an essential component of the big-data ecosystem to collect
batching data and streaming data.
4.2.2. Storage
Data storage is critical to big data applications in smart manu-
facturing. As the manufacturing industry increasingly benefits from the
use of big data, it is of importance to store more data [4]. Various
applications require different storage technologies to provide different
features, which are file system and databases. Although both technol-
ogies can store structured, semi-structured and unstructured data, they
have some differences that file system is suitable to store data-at-rest or
Y. Cui, et al.
unstructured data such as files, search and compile files manually.
Database is suitable to store data-in-motion, semi-structured or struc-
tured data, faster query data automatically.
There are three types of databases: RDBMS, NoSQL and NewSQL
databases. RDBMS database has been used in manufacturing for general
purpose applications for decades such as SQL Server, Oracle, MySQL.
Whereas, RDBMS is unable to address the challenges of big data's 3Vs
(Volume, Velocity and Variety) for storage and query [32]. NoSQL and
NewSQL databases can provide almost unlimited scalability and faster
query capability for industrial big data. Fig. 4 illustrates that the in-
creased use of a NoSQL database is already happening in manu-
facturing. NoSQL is suitable for one kind of OLTP application, which
does not require consistent data all the time, but simple query and
frequent updates to data [168]. Hence, manufacturing could use NoSQL
databases for real-time big data analytics such as quality monitoring
and prediction. Moreover, manufacturing could benefit from using the
four types of NoSQL data models (Key-value, Wide column, Document,
Graph) to easily manage semi-structured data (XML and JSON) [43].
The widely used NoSQL databases are Redis, HBase, Cassandra, Mon-
goDB and Neo4j. Since NoSQL databases were not designed to meet
data consistency, they are not suitable for some OLTP applications,
where data consistency needs to be guaranteed anytime [169]. There-
fore, NoSQL should not be used in an environment of many operations
and controllers that are present since the control data may be incon-
sistent. The NewSQL solution is used to provide relational query (SQL)
and data consistency all the time for an OLTP application. Manu-
facturing could use NewSQL for the scenarios requiring consistent data
all the time (finance data in ERP, inventory data in SCM, control signal
in SCADA systems). Some examples of NewSQL are VoltDB, Clustrix
and NuoDB [43]. However, NewSQL focuses on relational data, which
may not fully support unstructured data for big data analytics.
Unlike database solutions, in which the data is structured as a data
model for data consumers’ demands, file systems do not need a data
model to store unstructured data. Hadoop Distributed File System
(HDFS) could store petabytes of data with a redundantly low-cost
method [30]. However, querying data in HDFS is much slower than the
speed in databases. HDFS could be utilized as a central data storage to
retain all the data in manufacturing. Therefore, selection of the correct
storage solutions is required for big data applications in manufacturing.
4.2.3. Computation
Computation is the foundation of implementing big data applica-
tions. Three types of computation engines are available to manu-
facturing: batching, micro-batching and streaming. MapReduce in
Hadoop provides the batching method to process a petabyte level of big
data by less memory usage and cannot provide real-time analytics
[170]. Spark is a micro-batching processing engine, which provides
near real-time computation with more memory resources than Ma-
pReduce. Flink and Storm are real-time streaming engines to process
small volumes of data [171]. The differences and application scenarios
of these computation engines have been discussed in Section 3.3. Based
on the analysis of these computation engines and outcomes of Fig. 4,
there would be more Spark-based big data solutions on the factory floor
from device to production planning in the near future. The proposed
solutions could use Spark to replace Hadoop to get faster outcomes.
4.2.4. Analytics
Big data analytics is intended to extract information from collected
big data. Big data analytics includes two types of analysis: 1) data
mining and machine learning algorithms (clustering, regression,
Bayesian networks, artificial neural networks (ANN), deep learning. 2)
On-Line Analytic Processing (OLAP).
Big data analytics tools of the first type are identified as MLlib and
Scikit-learn for machine learning [25,109], Spark R for high-level sta-
tistical analysis [73], Tensorflow and CNTK for deep learning [105].
Analytics tools have many potential use cases in manufacturing, such as
12
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
machine vision for robotics [105], speech recognition for alarm and
security, image processing for quality control [87]and O&M [82].
Manufacturers can save time to develop algorithms from scratch by
using these tools in their big data platforms, such as image recognition
for product quality control [87]. Manufacturers also benefits from new
tools such as CaffeOnSpark [105], which is a deep learning framework
widely used for autonomous driving in the automotive industry [105].
Because Caffe does not work on Spark clusters, it is challenging to meet
the two strengths of better data analytics algorithm and faster big data
computation at the same time. CaffeOnSpark could address the issue to
work on Spark and Hadoop clusters for Caffe applications.
OLAP is an approach to analyse large multidimensional datasets for
complex business analytics, such as BI reporting, Decision Support
System (DSS), and CRM in manufacturing. Analytics tools of OLAP are
Apache Hive [172] and Apache Kylin [173]. Hive is one utility of Ha-
doop ecosystem which is used as a data warehouse. Kylin can query a
large volume of data faster than Hive.
By analyzing the collected big data, data analytics tools can effi-
ciently extract timely information to manufacturers to make decisions.
It is challenging to the decision makers to apply their experience and
knowledge on the new circumstances. Their experience is acquired
under the previous circumstance, which may be different from the
current one and their knowledge may be out of date. With the given
data and big data analytics tools, manufacturers can analyze historic
data, discover new knowledge, build actionable intelligence to make
data-driven decisions. It would happen in some areas of different sys-
tems feed with new data such as production planning with streaming
real-time IoT data.
4.2.5. Visualization
Manufacturing systems require various visualization methods to
present analytics results, such as interactive dashboard, reporting,
graph, document. Most of the proposed solutions did not provide a tool
to construct visualization. As the Python language is widely used in big
data analytics, it should be convenient to use Python plotting library
(Matplotlib) to present big data analytic results by data scientists in
manufacturing. However, Matplotlib works on a command line inter-
face (CLI), which is not user-friendly to business users with less pro-
gramming experience. The reviewed solutions could benefit from vi-
sualization tools such as Zeppelin, Tableau, and D3.Js [76]. Zeppelin
provides interpreters with big data tools (HBase, Cassandra, HDFS,
Spark, Flink etc.) and supports multi-agents. Manufacturers could use
Zeppelin in their existing big data platform with less developing effort.
D3.Js provides more complex visualization templates than Zeppelin.
D3.Js is a JavaScript library for producing dynamic, interactive data
visualizations in the web browser. It is suitable for applications, which
require dynamic monitoring and control such as MES, SCADA, O&M,
Quality control, safety and security systems [126]. D3.Js has less sup-
port to produce a report, which is not suitable for business analyses.
Tableau and QlikView are commercial data visualization software fo-
cusing on BI and support many common databases. . Big data and vi-
sualization can help bring data together and show the value of big data
in meaningful ways. AI helps to make automatic decisions, and visua-
lization helps to make manual decisions. As a result, AI and human can
collaboratively make data-driven decisions based on the actionable
intelligence generated by big data or mined by machine learning.
4.2.6. Workflow and dataflow
Workflow and dataflow components provide efficient approaches to
manage workflows for business and processes to data management.
Two workflow tools are found in the review papers: Oozie [116] and
Kelper [63]. Oozie is specialized in managing Hadoop jobs, which
would be suitable for a manufacturer with Hadoop ready platform
[174] . Kepler provides a convenient method to share workflows with
other business users in cloud manufacturing [63]. However, Kepler
does not support horizontal scalability to store large volume data.
Y. Cui, et al.
Another two workflow tools Wings and IBM InfoSphere have not been
found through this review [175]. Wings/Pegasus provides a more in-
telligent method to construct and execute workflow for users by opti-
mizing the workflow automatically. For parallel executing of work-
flows, Wings has to work with a resource management framework
(Pegasus), which may be a constraint to manufacturers with Hadoop-
based platform since Hadoop uses Yarn as its cluster resource manager.
IBM InfoSphere is a commercial enterprise-ready framework, which is
suitable for inexperienced users. It can work with Hadoop and other
IBM software together. InfoSphere is not open source, which is unable
to build add-on functions. For the dataflow component, researchers had
to construct their dataflow component from scratch. Usually, there is a
long learning curve for beginners to go through trial-error processes.
Apache NiFi is an efficient tool to construct dataflow and data in-
tegration with an interactive GUI. One paper uses NiFi to collect
streaming data in process industry [123]. NiFi is initially developed and
used by the National Security Agency of the United States (NSA), which
has been verified in the real application environment. Through this
study, there is limited work reported in this area. Smart manufacturing
needs the workflow and dataflow components to get data automatically
processed among various manufacturing systems. If assisted by AI and
optimization, they could significantly improve the efficiency of work-
flow and dataflow in manufacturing such as production planning and
scheduling.
4.2.7. Data management
Data management would make big data platform more feasible for
manufacturers. Data management focuses on data governance, meta-
data management, data modeling, data quality management, master
data management, data integration and knowledge management.
Although data management is highly related to the traditional IT area,
it provides a holistic management method to meet the requirements of
enterprise compliance, data policy, data lifecycle.
Some data management tools are available to the big data eco-
system of manufacturing:
• Data lifecycle management is based on enterprises’ policy to manage
data lifecycle from creation, storage, obsolescence to delete. Apache
Falcon includes the functions of data retention (persistency), data
replication for disaster recovery, aggregation and archive on
Hadoop [176]. Falcon is beneficial to manufacturing research. For
example, each acoustic experiment of aerospace engine collects
hundreds of gigabytes of audio data through over a hundred sensors.
The traditional method uses a single disk to store them, which is
costly to manage with high risk of disk failure or data loss. Falcon
could manage the data lifecycle with a friendly user interface.
• Data governance provides an enterprise policy approach to manage
data availability, usability, and integrity. Apache Atlas is the
Hadoop ecosystem tool for audit, lineage and service level agree-
ment (SLA), which is used to apply agile enterprise compliance
through consistent metadata management across the big data eco-
system [177].
• Data authorization manages users’ privilege to access sensitive data.
Apache Sentry is utilized to authorize data and metadata on Hadoop
clusters based on the roles of people in manufacturing [121].
• Data quality management is crucial to the outcomes of data ana-
lyses. Table 2 presents that data quality is a challenging data issue to
manufacturing. Apache Griffin maintains data quality by auto-
mating data profiling and validation on Hadoop and Spark [177].
Apache Griffin improves data quality by pre-processing data auto-
matically from various data sources, which reduces the amount of
data analyst's time to prepare data for analyzing.
• Data integration has two approaches to address data silos issue
[178]: Data Warehouse (DW) and Data Lake (DL). DW is the tradi-
tional approach that integrates data from various data sources into a
central data store with a predefined extract-transform-load (ETL)
13
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
method (schema-on-read) [176]. As data volume increases, dis-
tributed DW solution (Hive) and associated ETL tool (Apache Pig) is
available to manufacturing for business applications such as BI re-
porting, DSS. However, the predefined ETL method is not flexible
and expensive to build before this finally used by data consumers.
Data Lake is a new solution of data integration, which is defined as
central storage to store any data (sizes, types, rates) with the raw
formats in an enterprise [190]. DL is more suitable to data con-
sumers for an ad-hoc query of the data, which is undefined until
issuing the query (schema-on-read) [179]. HDFS is a popular DL tool
to store extensive unstructured data in manufacturing (video, audio,
image) [105]. However, DL probably becomes a “data swamp”
without practical data management tools (Falcon, Sentry, Atlas etc.)
[176].
• Semantic KM includes a series of operations for including, creating,
classifying, sharing, using information and knowledge in manu-
facturing [180]. Many technologies of the semantic web are utilized
to manage knowledge in manufacturing such as application of on-
tology web language (OWL) in SCADA [134], Resource Description
Framework (RDF), RDF query language (SPARQL) in SCM [91], RDF
database(Jena) in KM [23].
Data management tools include broad areas to address the veracity
and value of big data issues in manufacturing. The tools are still de-
veloping since most of the tools are Hadoop-based. As manufacturing
begins to focus on velocity of big data computation, some new data
management tools based on faster computation engines would be de-
veloped in the near future such as Spark and Flink. The big data
management tools are still manually operated by data stewards, which
would be challenging to manufacturers when various massive data is
ingested into the big data platform, and different roles of users apply to
use them. New methods of big data management could address this
issue by using algorithms such as rule-based, machine learning–based
or hybrid of both.
4.2.8. Infrastructure and deployment model
The infrastructure and deployment model are the foundation of big
data applications in manufacturing. There are two types of cloud
computing infrastructures:
1 General purpose commodity computer cluster (Hadoop, Spark) to
process completely parallel computing problems such as computing
massive sensor data separately or responding millions of users’ re-
quests (Facebook);
2 High-Performance Computing (HPC) cluster provides faster com-
puting speed with dedicated hardware.
Hence, HPC outperforms others in processing highly dependent data
computing such as complex modelling and simulation workload in
manufacturing. However, the disadvantages of HPC are that its in-
vestment is enormous, and its utilization rate is low. Some solutions
address its low utilization by moving HPC to cloud [182]. Some IAAS
providers offer public cloud HPC (AWS) and hybrid cloud HPC service
(Microsoft Azure).
Manufacturing enterprises have different perspectives on economic,
security, the privacy of big data platforms. Four deployment models
(Public cloud, Private cloud, Hybrid cloud, Community cloud [19]) are
available to satisfy the requirements [183]. For example: in terms of
better privacy and reducing waste of computing resource, some plat-
forms could adopt a hybrid cloud model, which puts sensitive data on
private cloud and process insensitive data on public cloud. However,
Hybrid and Community cloud are not found by this review. To meet
diverse requirements of manufacturing enterprises, some deployment
technologies, including OpenStack and Docker [191], are used to
quickly deploy an agile software environment with different micro-
service frameworks and programming languages.
Y. Cui, et al.
All deployment models and infrastructure should be taken into
consideration in the big data ecosystem of manufacturing.
4.2.9. Cybersecurity
Cybersecurity is an essential component of big data platform to
protect data assets in manufacturing. As manufacturing becomes data-
driven, the standards and tools are required to secure data in the IT
architecture of manufacturers completely. Although some security
standards have been provided in manufacturing systems such as SCADA
[192], the standards cannot fully address the challenges of SCADA on
the Internet [193]. Traditional control systems are vulnerable to un-
authorized attacks from the Internet since they are designed as close
systems with few capabilities of cybersecurity [194]. Because big data
platform integrates the physical space and the cyber space closely, the
risk of cybersecurity could quickly escalate to the physical system in
manufacturing. If unauthorized people manage the critical equipment
or information, it will bring irreversible disaster to manufacturing such
as economic loss, personal safety.
4.3. Question 3: how can we harness the capabilities available in the big
data ecosystem to drive research innovations in manufacturing?
Big data ecosystem is the comprehension of massive functional
components with various enabling tools. Capabilities of the big data
ecosystem are not only about computing and storing big data, but also
the advantages of its systematic platform and potentials of big data
analytics. Hence, according to proposed solutions of reviewed literature
and big data capabilities, the maturity of big data ecosystem application
is categorized into three stages:
Stage 1: proposing a big data framework and platform;
Stage 2: harvesting cloud computing capacity for big data com-
puting and storage;
Stage 3: analysing big data with various algorithms for the appli-
cations (prediction, fault detection, optimisation etc.).
Table 3 presents the allocation of proposed solutions by these three
stages. MES, SCADA and O&M have been studied within the three
stages. However, big-data based solution is missing in some manu-
facturing systems. For example, no general big data framework has
been identified in CAx (Stage 1); no practices of Stage 2 and Stage 3 are
found in PLM, SCM, BI and AM. Available capabilities of big data
ecosystems could drive research innovation in various big data appli-
cations (Fig. 5) of the immature manufacturing areas.
Manufacturing could benefit from new development and deploy-
ment methods of the big data ecosystem. Because many tools are fast
changing with highly frequent updates, it makes the design of the big
data platform dramatically challenging. Most of the proposed solutions
in manufacturing constructed the platform from scratch by installing
and testing every tool step by step. Researchers spent much time pre-
paring the software but not focusing on programming and big data
analytics. Several popular vendors offer their Hadoop distributions to
mitigate the issue, such as Cloudera (CDH), Hortonworks (HDP, HDF),
MapR, IBM (Infosphere BigInsights), Microsoft (HD Insight), Pivotal HD
[195]. Because the tools and versions were tested in the distributions,
they are ready to be used by researchers and enterprises in manu-
facturing. Moreover, the packages could be easily deployed and shared
with virtualization technology (VM and container) [191]. However,
developers need to evaluate several conditions to select the suitable
distribution such as open source, pricing, customer support, sizes of
community and so on [195].
4.4. Question 4: what are the future directions of big data applications in
manufacturing?
Manufacturing could benefit from fast developing big data
14
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
technology with new and matured tools. Hence, the full potential of big
data has not been discovered in manufacturing. Some potential direc-
tions are proposed for the future work of interested researchers:
4.4.1. Modeling and simulation
Modeling and simulation will naturally play an important role in
extracting value from data once the volume of data is available [4].
Digital twin and digital thread are the two essential methods with two
important tasks: collecting real-time data from manufacturing input
and output devices such as CNC machine and sensors and building up
the real-time simulation model with the collected data. These two tasks
require considerable computing resources and data storage to process
streaming data from manufacturing devices. Researchers commonly use
Matlab or some scientific software to develop the selected algorithm to
simulate the models [196]. As one simulation model may be im-
plemented on a single computer, many models are required to design,
and manually implemented on computers. It is a challenging and on-
erous task to manufacturers to convert the Matlab simulation models to
executive programs on the selected cloud computing engine. However,
from the reviewed articles, simulation is rare in the big data based
solutions. The ongoing practice is that of a digital twin, which provide a
real-time, bi-directional management between a physical object and its
digital object [36]. It is essential to simulate the digital twin model with
parameters across different domains (e.g. product, process and logis-
tics) in order to predict its dynamic performance, such as Virtual/
Augmented Reality [4]. For example, simulation of the product model
with predefined parameters includes FEA analysis (ANSYS). However,
the parameters from other domains (processes and logistics) are not
accounted for in the FEA simulation. One potential research direction is
customized products by integrating End-Of-Life data from PLM into
CAx software. It could improve the product's impact during its EOL
phase, such as Prognostics and health management (PHM) [184]).
Another direction is that simulation of digital twin models with real-
time data could be used for predictive maintenance. Moreover, simu-
lation can provide integral data to data scientists or business users to
design machine learning algorithms. Because Matlab simulation man-
ager is available to simulate multiple models on the public cloud
(Azure, AWS), more case studies of methods above could be im-
plemented to verify the performance of big data tools for digital twin
and digital thread. Since Matlab is commercial software, unlike most of
the software in the big data ecosystem that is open source and free to
use, users have to purchase the specific licenses to run the models on
the cloud. AnyLogic is also a very popular simulation software which
provides cloud-based simulation recently. Researchers could use Any-
Logic Cloud to deploy and verify their digital twin models.
Simulations with these large volumes of data require enormous
computing, storage and communication resources, and cloud com-
puting is well placed to satiate these. Another disruptive technology is
Quantum computing, which can provide unlimited resources to process
and store the data [197]. However, the technology is still under de-
velopment and still has limitations, such as fault tolerance and error
correlation [185].
Future research directions could be:
• Develop big data tools to convert simulation models from scientific
software to implement on public cloud or private cloud;
• Using a simulation method to generate testing and training data for
machine learning at planning and decision-making processes;
• Using general-purpose cloud computing cluster to simulate FEA at
product design.
4.4.2. Connectivity and interoperability
Big data ecosystem complements of existing manufacturing ap-
proaches to have connectivity and interoperability. With the aim of
systems integration and collaboration for smart manufacturing, the
basis is that data and information have to be timely collected, correctly
Y. Cui, et al.
formatted, analyzed and exchanged among the systems. MTConnect
and OPC-UA are both data collection approaches in manufacturing.
MTConnect focuses on device level and control level by monitoring
CNC machine tools using a predefined consistent data model, data
format and definition, which matches different vendors' machine tools.
OPC-UA focuses on SCADA, MES and ERP by using a generic data
model which can flexibly match more industrial devices with additional
configuration effort. There are three challenging issues of both ap-
proaches. One issue is the performance limitation of OPC-UA which the
CPU of OPC-UA server is identified as the main bottleneck in produc-
tion [198]. Since both approaches are implemented on edge devices at
the factory floor, the capabilities of data processing cannot be flexibly
scaled up. The servers of both approaches have to be replaced with
better performing hardware, while more data sources are connecting to
the systems such as IoT devices. Another issue is interoperability of the
systems using both approaches since there is no common ontology on
the top of both information models. The last issue is the capabilities of
data analysis and information exchange between both systems and
other systems such as SCM, PLM, and CRM. Although some solutions
are proposed to exchange both data formats and other formats (such as
MTConnect and IEEE 1451 [118], OPC-UA and IoT [117], OPC-UA and
AutomationML [98]), more solutions are required for data exchange of
massive data formats between these two systems and others systems as
shown in Table 1.
Four existing big data ingestion tools (Sqoop, Flume, Kafka and
Storm) and a new tool (NiFi) could complement the weaknesses of both
approaches. All five big data tools can scale out the capability of data
processing by adding new hardware to the clusters without replacing
the old ones. There are some differences between the five tools. The
first difference is that Sqoop and Flume are two tools of Hadoop eco-
system; while Kafka and Storm are not dedicated to Hadoop. Secondly,
data consumers require to pull data with Kafka, whereas Flume pushes
data to consumers. Thirdly, Kafka provides better fault-tolerance and
scalable than others. Kafka provides event duplication, which means
other nodes continuously make data available when one node is failed.
Compared with Kafka, Storm, Flume and NiFi do not provide event
duplication, which is theoretically not suitable for application of critical
missions such as safety, security, and finance in manufacturing. Storm,
Flume and NiFi require less developing effort to work with Hadoop.
Interoperability has two main issues: data format and data quality.
With the aim of interoperability among systems on BDA platform, data
needs to be correctly formatted with good quality before exchange
happens. Table 1 demonstrated various data formats in manufacturing
systems. In terms of data formats, there are two types of data: data with
schema (XML with given XSD/DTD) and data without schema such as
XML without given XSD/DTD, JSON and unstructured data (document,
report, email). Since data with schema define the schemas in standard
schema files (XSD/DTD: PLM XML), data is can be exchanged auto-
matically by mapping elements of each other's schemas. It is challen-
ging to automatically transform data without schema to an intended
schema since there is no mapping between both sides. One solution may
be using natural language processing (NLP) to process human readable
documents. BDA platform also has to support new file formats such as
AMF and 3MF [199].
Issues of data quality in Table 2 includes missing value, noise data
or anomaly data, uncertainty, data outlier, data correlation, timing and
synchronization. Data transformation can address these issues to extract
data to correct timely information. No generic transformation tool is
identified in the reviewed papers which can be used as “One size fit all”
tool for all the applications in manufacturing. Because manufacturers
use various manufacturing systems with different data characterizes,
there are many combinations of data issues which require massive data
transformation tools to address. It is challenging for manufacturers to
develop customized big data transformation tool for every specific
scenario. Without addressing these data issues, correct information
cannot be extracted and exchanged among systems. Systems integration
15
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
and collaboration cannot be achieved for smart manufacturing.
Future research directions in this area could be:
• Review the availability and feasibility of data collection tools in
various manufacturing scenarios;
• Develop a generic data transformation solution with big data tech-
nologies to exchange data of manufacturing systems;
4.4.3. Standardized big data platform design
Standardization of big data platform design improves the feasibility
of enterprise-ready solutions in manufacturing. Because some essential
components are missing in the proposed solutions, it is likely to increase
the difficulty to apply them to manufacturing. Missing components
would be insufficient to apply big data applications to smart manu-
facturing. However, there is no standard approach to design big data
platform in manufacturing. The reason may be that different profiles of
manufacturing enterprises have varieties of system requirements for big
data applications.
A systematic assessment method is required to analyze the limita-
tions and strengths of the proposed big data solution in the various
manufacturing systems. Data issues are not entirely assessed, such as
latency of data transmission among clusters, data quality and data
format exchange. Therefore, it is essential to provide a standard ap-
proach to design big data platform with related assessment method to
manufacturing.
4.4.4. Real time big data analytics
Big data analytics go deeper from batch analysing to real-time
streaming analyzing in manufacturing. On the one hand, streaming big
data analysis is considered as a high research requirement in manu-
facturing [187]. One the other hand, enabling technologies are chan-
ging from non-real time analytics to real time. Fig. 4 illustrates that the
streaming computation engine Spark is becoming a popular tool than
the traditional batching engine (MapReduce of Hadoop). However, it
also shows that Hadoop still receives more focus than Spark in manu-
facturing from Fig. 4. Batch processing is not able to provide real-time
analytics response such as real time monitoring, dynamic scheduling
and planning on systems of workshop floor (SCADA, MES). Micro-
batching and streaming engines (Spark, Storm, Flink) can provide real-
time big data analytics of streaming data [26]. Spark could take ad-
vantage of batching and streaming to replace MapReduce engine.
Moreover, compared to Storm and Flink, Spark has more powerful
analytics tools such as SparkSQL, Spark R, GraphX, and MLlib.
However, Storm and Flink outperform Spark in real-time concerns.
Hence, with the objective of streaming and analytics, it is necessary to
use Storm or Flink with Spark together. The issue is that it requires
more development effort to work on several computation engines.
Apache Beam provides a uniform abstraction layer to run these real-
time engines at the execution layer [44]. Although it has not been used
in manufacturing, researchers in manufacturing could focus on analy-
tics logics without spending time on learning various usages of engines.
Another fact of big data analytics forwarding to real-time is that data
warehousing tools (Pig, Hive) of BI used MapReduce to batch proces-
sing large datasets, now that their latest versions support Spark engine.
4.4.5. Cybersecurity in manufacturing
Cybersecurity will continuously challenge manufacturing since se-
curity standards are still not available in some system such as SCM [49].
Recently, NIST published a framework to improve cybersecurity on
critical infrastructures [188]. It is envisioned that there would be more
developments in cybersecurity tools based on the new standard. No big
data tool of cybersecurity was found in all reviewed papers, which is
likely a promising research direction. Manufacturing could benefit from
the following new security tools. Firstly, Apache Metron is an en-
terprise-ready real-time big data security tool, which is used by Telstra
Company [200]. Secondly, Apache Ranger provides security
Y. Cui, et al. Robotics and Computer Integrated Manufacturing 62 (2020) 101861

administration management on Hadoop clusters. Thirdly, Apache Knox
provides gateway service to access Hadoop clusters. Future direction in
this area is to explore the capability of big data cybersecurity tools on
critical systems in manufacturing such as safety, security and SCADA.
Vulnerability in both the physical and cyberspace of the manufacturing
systems must be identified and protected. The potential risk and da-
mages warrant high priority of future research in this direction.
5. Conclusions and future work
This paper systematically reviews the state of art of big data re-
search in manufacturing to evaluate the capabilities of big data eco-
system and requirements of smart manufacturing. Six key drivers of big
data ecosystem are identified for smart manufacturing, which are
system integration, data, prediction, sustainability, resource sharing
and hardware. Afterwards, the nine essential components of big data
ecosystem are presented to design a feasible big data solution to man-
ufacturing enterprises. These are data ingestion, storage, computing,
analytics, visualization, management, workflow, infrastructure and se-
curity. The evaluation reveals that there is no enterprise-ready big data
solution in the reviewed literature.
It is important to note that some research areas have received less
attention from the manufacturing community such as PLM, CAx, ERP
and SCM. Many big data utilities are applicable to these areas, which
could drive research innovation.
Regarding future work, there are five promising directions: mod-
eling and simulation, connectivity and interoperability, standardized
big data platform design, real-time big data analytics and cybersecurity.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.rcim.2019.101861.

Appendices

Tables A1 and A2.

Table A1
Search strings of four citation databases.
Search strings Number of papers

Scopus TITLE-ABS-KEY (("hadoop") OR ("spark") OR ("Storm") OR ("Flink") OR ("SQL") OR ("nosyl") OR ("time-series database") OR ("opc-usa") OR 228
("MTconnect")) AND (TITLE-ABS-KEY ("manufacturing") OR TITLE-ABS-KEY ("Industry 4.0”) OR TITLE-ABS-KEY ("Smart manufacturing") OR
TITLE-ABS-KEY ("Digital twin") OR TITLE-ABS-KEY ("Digital thread")) AND NOT (TITLE-ABS-KEY ("construction") OR TITLE-ABS-KEY
("healthcare") OR TITLE-ABS-KEY ("oil") OR TITLE-ABS-KEY ("energy") OR TITLE-ABS-KEY ("Agriculture"))
IEEE Xplore (("Author Keywords":hadoop OR "Author Keywords":spark OR "Author Keywords":storm OR "Author Keywords":flink OR "Author 63
Keywords":sql OR "Author Keywords":nosql OR "Author Keywords":"time-series database" OR "Author Keywords":"opc-ua" OR "Author
Keywords":"mtconnect") AND ("Abstract":manufacturing OR "Abstract":"Industrial 4.0” OR "Abstract":"industrial automation" OR
"Abstract":"smart manufacturing" OR "Abstract":"digital twin" OR "Abstract":"digital thread"))
ASME (hadoop, spark, storm, flink, sql, nosql, "time series database", "opc-ua", mtconnect) AND (manufacturing, "industry 4.0”, "industrial 25
automation", "smart manufacturing", "digital twin", "digital thread")
ACM ((keywords.author.keyword:(hadoop, spark, storm, flink) AND content.ftsec:(apache)) OR keywords.author.keyword:(sql, nosql, "time series 22
database", opc-ua, mtconnect)) AND recordAbstract:(manufacturing, "industrial automation", "smart manufacturing", "digital twin", "digital
thread")

Table A2
Category distribution of reviewed articles.
Categories Systems Reference

Product CAD/CAE/CAPP/CAM [55,53,56,125]

PLM [58,59]
Production ERP [126,62,62,97,94]
MOM/MES [25,65,127,128,69,66,116,6]
SCADA/DCS/HMI [137,73,71,39,134,201,133,131,130,129]
IIoT/CNC/Robot [166,202,78,77,140]
O&M [138,24,83,139,161,156,157,142,82]
QMS [85,88,109,87]
Safety [89,203]
Business SCM/CRM/BI/AM [93,90,181]
ICT architecture CPS/CM/ICT [64,146,68,160,153,99,132,101,143,75,148,95,149,204,96,76]
Data analytics/DM [105,121,205,102,152,206,151,207,208]
KM [164,23,106]

References Intelligence). Johannes, Recommendations for implementing the strategic in-

[1] S. de Treville, M. Ketokivi, V. Singhal, Competitive manufacturing in a high-cost
environment: introduction to the special issue, J. Oper. Manag. 49–51 (2017) 1–5,
https://doi.org/10.1016/j.jom.2017.02.001.
[2] H. (Deutsche P.A. Henning, Kagermann (National Academy Of Science and
Engineering). Wolfgang, Wahlster (German Research Center for Artificial
itiative INDUSTRIE 4.0, Final Rep. Ind. 4.0 WG. (2013) 82.
[3] Y. Liao, F. Deschamps, E. de, F.R. Loures, L.F.P. Ramos, Past, present and future of
industry 4.0 - a systematic literature review and research agenda proposal, Int. J.
Prod. Res. 55 (2017) 3609–3629, https://doi.org/10.1080/00207543.2017.
1308576.
[4] A. Kusiak, Smart manufacturing, Int. J. Prod. Res. 56 (2018) 508–517, https://doi.
org/10.1080/00207543.2017.1351644.

16
Y. Cui, et al.
[5] B. (Serm) Kulvatunyou, N. Ivezic, V. Srinivasan, On architecting and composing
engineering information services to enable smart manufacturing, J. Comput. Inf.
Sci. Eng. 16 (2016) 031002, , https://doi.org/10.1115/1.4033725.
[6] B.W. Jeon, J. Um, S.C. Yoon, S. Suk-Hwan, An architecture design for smart
manufacturing execution system, Comput. Aided. Des. Appl. 4360 (2016) 1–14,
https://doi.org/10.1080/16864360.2016.1257189.
[7] X. Xu, From cloud computing to cloud manufacturing, robot, Comput. Integr.
Manuf. 28 (2012) 75–86, https://doi.org/10.1016/j.rcim.2011.07.002.
[8] J. Li, F. Tao, Y. Cheng, L. Zhao, Big Data in product lifecycle management, Int. J.
Adv. Manuf. Technol. 81 (2015) 667–684, https://doi.org/10.1007/s00170-015-
7151-x.
[9] J. Wang, Q. Chang, G. Xiao, N. Wang, S. Li, Data driven production modeling and
simulation of complex automobile general assembly plant, Comput. Ind. 62 (2011)
765–775, https://doi.org/10.1016/j.compind.2011.05.004.
[10] S. Yang, B. Bagheri, H.-A. Kao, J. Lee, A unified framework and platform for de-
signing of cloud-based machine health monitoring and manufacturing systems, J.
Manuf. Sci. Eng. 137 (2015) 040914, , https://doi.org/10.1115/1.4030669.
[11] H. Sequeira, P. Carreira, T. Goldschmidt, P. Vorst, Energy cloud: real-time cloud-
native energy management system to monitor and analyze energy consumption in
multiple industrial sites, Proc. - 2014 IEEE/ACM 7th Int. Conf. Util. Cloud Comput.
UCC 2014, 2014, pp. 529–534, , https://doi.org/10.1109/UCC.2014.79.
[12] R.Y. Zhong, S.T. Newman, G.Q. Huang, S. Lan, Big Data for supply chain man-
agement in the service and manufacturing sectors: challenges, opportunities, and
future perspectives, Comput. Ind. Eng. 101 (2016) 572–591, https://doi.org/10.
1016/j.cie.2016.07.013.
[13] H.M. Chen, R. Schutz, R. Kazman, F. Matthes, Amazon in the air: innovating with
big data at Lufthansa, Proc. Annu. Hawaii Int. Conf. Syst. Sci. 2016-March, 2016,
pp. 5096–5105, , https://doi.org/10.1109/HICSS.2016.631.
[14] M. Helu, T. Hedberg, A. Barnard Feeney, Reference architecture to integrate
heterogeneous manufacturing systems for the digital thread, CIRP J. Manuf. Sci.
Technol. 19 (2017) 191–195, https://doi.org/10.1016/j.cirpj.2017.04.002.
[15] F. Tao, F. Sui, A. Liu, Q. Qi, M. Zhang, B. Song, Z. Guo, S.C.Y. Lu, A.Y.C. Nee,
Digital twin-driven product design framework, Int. J. Prod. Res. 7543 (2018)
1–19, https://doi.org/10.1080/00207543.2018.1443229.
[16] L. Monostori, B. Kádár, T. Bauernhansl, S. Kondoh, S. Kumara, G. Reinhart,
O. Sauer, G. Schuh, W. Sihn, K. Ueda, Cyber-physical systems in manufacturing,
CIRP Ann. Manuf. Technol. 65 (2016) 621–641, https://doi.org/10.1016/j.cirp.
2016.06.005.
[17] F. Bonomi, R. Milito, P. Natarajan, J. Zhu, Big Data and internet of things: aA
roadmap for smart environments, 2014. doi:10.1007/978-3-319-05029-4.
[18] J. Gubbi, R. Buyya, S. Marusic, M. Palaniswami, Internet of things (IoT): a vision,
architectural elements, and future directions, Futur. Gener. Comput. Syst. 29
(2013) 1645–1660, https://doi.org/10.1016/j.future.2013.01.010.
[19] P. Mell, T. Grance, The NIST definition of cloud computing, NIST Spec. Publ. 145
(2011) 7, https://doi.org/10.1136/emj.2010.096966.
[20] B.V. Dhar, Data science and prediction, Commun. ACM 56 (2013) 64–73.
[21] A.J.C. Trappey, C.V. Trappey, U. Hareesh Govindarajan, A.C. Chuang, J.J. Sun, A
review of essential standards and patent landscapes for the internet of things: a key
enabler for industry 4.0, Adv. Eng. Informatics. 33 (2017) 208–229, https://doi.
org/10.1016/j.aei.2016.11.007.
[22] R. Kosara, C. Healey, Visualization viewpoints: data, information and knowledge
in visualization, Comput. Graph. …. (2003), http://ieeexplore.ieee.org/xpls/abs_
all.jsp?arnumber=1210860.
[23] S. Wang, J. Wan, D. Li, C. Liu, Knowledge reasoning with semantic data for real-
time data processing in smart factory, Sensors (Switzerland) 18 (2018) 1–10,
https://doi.org/10.3390/s18020471.
[24] L. Spendla, M. Kebisek, P. Tanuska, L. Hrcka, Concept of predictive maintenance
of production systems in accordance with industry 4.0, SAMI 2017 - IEEE 15th Int.
Symp. Appl. Mach. Intell. Informatics, Proc. 2017, pp. 405–410, , https://doi.org/
10.1109/SAMI.2017.7880343.
[25] O. Morariu, C. Morariu, T. Borangiu, S. Răileanu, Manufacturing systems at scale
with Big Data streaming and online machine learning, Stud. Comput. Intell. 762
(2018) 253–264, https://doi.org/10.1007/978-3-319-73751-5_19.
[26] K. Nagorny, P. Lima-Monteiro, J. Barata, A.W. Colombo, Big Data analysis in smart
manufacturing: a review, Int. J. Commun. Netw. Syst. Sci. 10 (2017) 31–58,
https://doi.org/10.4236/ijcns.2017.103003.
[27] D. Reinsel, J. Gantz, J. Rydning, Data age 2025: tThe digitization of the world
from edge to core, 2018. https://www.seagate.com/files/www-content/our-
story/trends/files/idc-seagate-dataage-whitepaper.pdf.
[28] A. Gandomi, M. Haider, Beyond the hype: Big data concepts, methods, and ana-
lytics, Int. J. Inf. Manage. 35 (2015) 137–144, https://doi.org/10.1016/j.
ijinfomgt.2014.10.007.
[29] J. Dean, S. Ghemawat, MapReduce: simplified data processing on large clusters,
Communications of the ACM 51 (1) (2008) 107–113.
[30] K. Shvachko, H. Kuang, S. Radia, R. Chansler, The Hadoop distributed file system,
2010 IEEE 26th Symp. Mass Storage Syst. Technol. 2010, pp. 1–10, , https://doi.
org/10.1109/MSST.2010.5496972.
[31] J. Duda, Business intelligence and NoSQL databases, Inf. Syst. Manag. 1 (2012)
25–37.
[32] R. Cattell, Scalable SQL and NoSQL data stores, Acm Sigmod Rec. 39 (2010)
12–27.
[33] Y. Bao, L. Ren, L. Zhang, X. Zhang, Y. Luo, Massive sensor data management
framework in cloud manufacturing based on Hadoop, IEEE Int. Conf. Ind.
Informatics, 2012, pp. 397–401, , https://doi.org/10.1109/INDIN.2012.6301192.
[34] F. Tao, Y. Cheng, L. Da Xu, L. Zhang, B.H. Li, CCIoT-CMfg: cloud computing and
internet of things-based cloud manufacturing service system, IEEE Trans. Ind.
17
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
Informatics 10 (2014) 1435–1442, https://doi.org/10.1109/TII.2014.2306383.
[35] D. Wu, D.W. Rosen, L. Wang, D. Schaefer, Cloud-based design and manufacturing:
A new paradigm in digital manufacturing and design innovation, CAD Comput.
Aided Des. 59 (2015) 1–14, https://doi.org/10.1016/j.cad.2014.07.006.
[36] F. Tao, Q. Qi, A. Liu, A. Kusiak, Data-driven smart manufacturing, J. Manuf. Syst.
(2018), https://doi.org/10.1016/j.jmsy.2018.01.006.
[37] M. Soualhia, F. Khomh, S. Tahar, Task scheduling in Big Data platforms: A sys-
tematic literature review, J. Syst. Softw. 134 (2017) 170–189, https://doi.org/10.
1016/j.jss.2017.09.001.
[38] Y. Demchenko, P. Grosso, C. De Laat, P. Membrey, Addressing Big Data issues in
scientific data infrastructure, Proc. 2013 Int. Conf. Collab. Technol. Syst. CTS
2013, 2013, pp. 48–55, , https://doi.org/10.1109/CTS.2013.6567203.
[39] J. Moyne, J. Samantaray, M. Armacost, Big Data capabilities applied to semi-
conductor manufacturing advanced process control, IEEE Trans. Semicond. Manuf.
29 (2016) 283–291, https://doi.org/10.1109/TSM.2016.2574130.
[40] Y. Demchenko, C. De Laat, P. Membrey, Defining architecture components of the
Big Data ecosystem, 2014 Int. Conf. Collab. Technol. Syst. CTS 2014, 2014, pp.
104–112, , https://doi.org/10.1109/CTS.2014.6867550.
[41] S. Landset, T.M. Khoshgoftaar, A.N. Richter, T. Hasanin, A survey of open source
tools for machine learning with big data in the Hadoop ecosystem, J. Big Data 2
(2015) 24, https://doi.org/10.1186/s40537-015-0032-1.
[42] S. Binani, A. Gutti, S. Upadhyay, SQL vs. NoSQL vs. NewSQL-a comparative study,
Commun. Appl. Electron. 6 (2016) 43–46 http://www.caeaccess.org/archives/
volume6/number1/binani-2016-cae-652418.pdf.
[43] K. Grolinger, W.A. Higashino, A. Tiwari, M.A.M. Capretz, Data management in
cloud environments: NoSQL and NewSQL data stores, J. Cloud Comput. (2013) 2,
https://doi.org/10.1186/2192-113X-2-22.
[44] M. Gökalp, K. Kayabay, M. Zaki, A. Koçyiğit, Big-Data data analytics architecture
for businesses: A comprehensive review on new open-source big-data tools,
Cambridge Service Alliance, 2017, pp. 1–35, , https://doi.org/10.13140/RG.2.2.
30306.84165.
[45] J. Wan, S. Tang, Z. Shu, D. Li, S. Wang, M. Imran, A.V. Vasilakos, Software-Defined
industrial internet of things in the context of industry 4.0, IEEE Sens. J. 16 (2016)
7373–7380, https://doi.org/10.1109/JSEN.2016.2565621.
[46] M.J. Koop, W. Huang, K. Gopalakrishnan, D.K. Panda, Performance analysis and
evaluation of PCIe 2.0 and quad-data rate InfiniBand, Proc. - Symp. High Perform.
Interconnects, Hot Interconnects, 2008, pp. 85–92, , https://doi.org/10.1109/
HOTI.2008.26.
[47] J. Cheng, L. Da Xu, W. Chen, F. Tao, C.-L. Lin, Industrial IoT in 5G environment
towards smart manufacturing, J. Ind. Inf. Integr. (2018), https://doi.org/10.1016/
j.jii.2018.04.001.
[48] P. Zikopoulos, C. Eaton, Understanding Big Data: aAnalytics for enterprise class
hadoop Hadoop and streaming data: aAnalytics for enterprise class hadoop
Hadoop and streaming data, 2011. https://www.immagic.com/eLibrary/
ARCHIVES/EBOOKS/I111025E.pdf.
[49] Y. Lu, K. Morris, S. Frechette, Current standards landscape for smart manu-
facturing systems, Natl. Inst. Stand. Technol. NISTIR 8107 (2016) 39, https://doi.
org/10.6028/NIST.IR.8107.
[50] S. Evdokimov, RFID and the internet of things: tTechnology, applications, and
security challenges, 2010. doi:10.1561/0200000020.
[51] D. Kiritsis, V.K. Nguyen, J. Stark, How closed-loop PLM improves knowledge
management over the complete product lifecycle and enables the factory of the
future, Int. J. Prod. Lifecycle Manag. 3 (2008) 54, https://doi.org/10.1504/IJPLM.
2008.019970.
[52] P. Buneman, Semistructured data, Proc. 16th Symp. Princ. Database Syst. 1997,
pp. 117–121, , https://doi.org/10.1145/263661.263675.
[53] Y. Xu, G. Chen, J. Zheng, An integrated solution—KAGFM for mass customization
in customer-oriented product design under cloud manufacturing environment, Int.
J. Adv. Manuf. Technol. 84 (2016) 85–101, https://doi.org/10.1007/s00170-015-
8074-2.
[54] S. Chen, C. Yin, X. Li, Implementation of MTConnect in machine monitoring
system for CNCs, Proc. - 2017 5th Int. Conf. Enterp. Syst. Ind. Digit. by Enterp.
Syst. ES 2017, 2017, pp. 70–75, , https://doi.org/10.1109/ES.2017.19.
[55] S.L. Chan, Y. Lu, Y. Wang, Data-driven cost estimation for additive manufacturing
in cybermanufacturing, J. Manuf. Syst. 46 (2018) 115–126, https://doi.org/10.
1016/j.jmsy.2017.12.001.
[56] S.K., K.C., H.Y., S.O. Yang, G-code conversion from 3D model data for 3D printers
on Hadoop systems, 2017 4th Int. Conf. Comput. Appl. Inf. Process. Technol. 2017,
pp. 1–4, , https://doi.org/10.1109/CAIPT.2017.8320709.
[57] S. Sierla, V. Kyrki, P. Aarnio, V. Vyatkin, Automatic assembly planning based on
digital product descriptions, Comput. Ind. 97 (2018) 34–46, https://doi.org/10.
1016/j.compind.2018.01.013.
[58] Y. Zhang, S. Ren, Y. Liu, T. Sakao, D. Huisingh, A framework for Big Data driven
product lifecycle management, J. Clean. Prod. 159 (2017) 229–240, https://doi.
org/10.1016/j.jclepro.2017.04.172.
[59] Y. Zhang, S. Ren, Y. Liu, S. Si, A big data analytics architecture for cleaner man-
ufacturing and maintenance processes of complex products, J. Clean. Prod. (2016),
https://doi.org/10.1016/j.jclepro.2016.07.123.
[60] D. Ramanujan, W.Z. Bernstein, M.A. Totorikaguena, C.F. Ilvig, K.B. Ørskov,
Generating contextual design for environment principles in sustainable manu-
facturing using visual analytics, J. Manuf. Sci. Eng. 141 (2018) 021016, , https://
doi.org/10.1115/1.4041835.
[61] T. Hedberg, A.B. Feeney, M. Helu, J.A. Camelio, Toward a lifecycle information
framework and technology in manufacturing, J. Comput. Inf. Sci. Eng. 17 (2017)
021010, , https://doi.org/10.1115/1.4034132.
[62] J.W. Wang, J.Y. Yang, J. Zhang, X.X. Wang, W. (Chris) Zhang, Big data driven
Y. Cui, et al.
cycle time parallel prediction for production planning in wafer manufacturing,
Enterp. Inf. Syst. 12 (2018) 1–19, https://doi.org/10.1080/17517575.2018.
1450998.
[63] X. Li, J. Song, B. Huang, A scientific workflow management system architecture
and its scheduling based on cloud service platform for manufacturing big data
analytics, Int. J. Adv. Manuf. Technol. 84 (2015) 119–131, https://doi.org/10.
1007/s00170-015-7804-9.
[64] M.Y. Santos, J. Oliveira e, C. Sá, F. Andrade, E. Vale Lima, C. Costa, B. Costa,
J. Martinho, A. Galvão, Big Data system supporting Bosch Braga industry 4.0
strategy, Int. J. Inf. Manage. 37 (2017) 750–760, https://doi.org/10.1016/j.
ijinfomgt.2017.07.012.
[65] S. Wang, C. Zhang, C. Liu, D. Li, H. Tang, Cloud-assisted interaction and nego-
tiation of industrial robots for the smart factory, Comput. Electr. Eng. 63 (2017)
66–78, https://doi.org/10.1016/j.compeleceng.2017.05.025.
[66] N.M. Khushairi, N.A. Emran, M.M. Mohd Yusof, Database performance tuning
methods for manufacturing execution system, World Appl. Sci. J. 30 (2014)
91–99, https://doi.org/10.5829/idosi.wasj.2014.30.icmrp.14.
[67] B.W. Jeon, J. Um, S.C. Yoon, S. Suk-hwan, An architecture design for smart
manufacturing execution system, 4360 (2017). doi:10.1080/16864360.2016.
1257189.
[68] M. Jirkovský, V., Obitko, Enabling semantics within industry 4.0, Int. Conf. Ind.
Appl. Holonic Multi-Agent Syst. 2017, pp. 39–52, , https://doi.org/10.1007/978-
3-642-40090-2.
[69] O. Sauer, Developments and trends in shopfloor-related ICT systems, IEEE Int.
Conf. Ind. Eng. Eng. Manag. 2015-Janua, 2014, pp. 1352–1356, , https://doi.org/
10.1109/IEEM.2014.7058859.
[70] B.E.I. Systron, D. Division, Managing configuration control in an automotive
sensor mass customization manufacturing product line, 2006 World Autom.
Congr. 2006, pp. 1–16.
[71] J.-J. Kim, D.-W. Lee, D.-B. Ko, S.-I. Jeong, J.-M. Park, An autonomic computing
based on big data platform for high-reliable smart factory, J. Eng. Appl. Sci. 12
(2017) 2662–2666 https://www.scopus.com/inward/record.uri?eid=2-s2.0-
85041023417&partnerID=40&md5=003f099909f142b4879dffa79001fe1c.
[72] P. Reboredo, M. Keinert, Integration of discrete manufacturing field devices data
and services based on OPC UA, IECON Proc. (Industrial Electron. Conf. 2013, pp.
4476–4481, , https://doi.org/10.1109/IECON.2013.6699856.
[73] A.R. Khan, H. Schioler, M. Kulahci, T. Knudsen, Big data analytics for industrial
process control, 2017 22nd IEEE Int. Conf. Emerg. Technol. Fact. Autom. 2017, pp.
1–8, , https://doi.org/10.1109/ETFA.2017.8247658.
[74] X. Ye, S.H. Hong, An AutomationML/OPC UA-based industry 4.0 solution for a
manufacturing system, IEEE Int. Conf. Emerg. Technol. Fact. Autom. ETFA. 2018-
September, 2018, pp. 543–550, , https://doi.org/10.1109/ETFA.2018.8502637.
[75] D. Mourtzis, E. Vlachou, A cloud-based cyber-physical system for adaptive shop-
floor scheduling and condition-based maintenance, J. Manuf. Syst. 47 (2018)
179–198, https://doi.org/10.1016/j.jmsy.2018.05.008.
[76] C. Toro, I. Barandiaran, J. Posada, A perspective on knowledge based and in-
telligent systems implementation in industrie 4.0, Procedia Comput. Sci. 60 (2015)
362–370, https://doi.org/10.1016/j.procs.2015.08.143.
[77] E. Poormohammady, J.H. Reelfs, M. Stoffers, K. Wehrle, A. Papageorgiou,
Dynamic algorithm selection for the logic of tasks in IoT stream processing sys-
tems, 2017 13th Int. Conf. Netw. Serv. Manag. CNSM 2017. 2018-Janua, 2018, pp.
1–5, , https://doi.org/10.23919/CNSM.2017.8256009.
[78] D. Mourtzis, E. Vlachou, N. Milas, Industrial Big Data as a result of IoT adoption in
manufacturing, Procedia CIRP 55 (2016) 290–295, https://doi.org/10.1016/j.
procir.2016.07.038.
[79] J. Moyne, J. Samantaray, M. Armacost, Big data emergence in semiconductor
manufacturing advanced process control, Adv. Semicond. Manuf. Conf. (ASMC),
2015 26th Annu. SEMI. 2015, pp. 130–135, , https://doi.org/10.1109/ASMC.
2015.7164483.
[80] D. Mourtzis, A. Vlachou, V. Zogopoulos, Cloud-based augmented reality remote
maintenance through shop-floor monitoring: A product-service system approach,
J. Manuf. Sci. Eng. 139 (2017) 061011, , https://doi.org/10.1115/1.4035721.
[81] Y. Busnel, N. Riveei, A. Gal, Y. Busnel, N. Riveei, A. Gal, A. Gal, FlinkMan :
aAnomaly detection in manufacturing equipment with Apache Flinkapache flink:
gGrand challenge, (2017). doi:10.1145/3093742.3095099.
[82] I. Yen, S. Zhang, F. Bastani, A framework for IoT-based monitoring and diagnosis
of manufacturing systems, Proc. - 11th IEEE Int. Symp. Serv. Syst. Eng. SOSE 2017,
2017, pp. 1–8, , https://doi.org/10.1109/SOSE.2017.26.
[83] S. Kang, W.T.K. Chien, J.G. Yang, A study for big-data (Hadoop) application in
semiconductor manufacturing, IEEE Int. Conf. Ind. Eng. Eng. Manag. 2016-Decem,
2016, pp. 1893–1897, , https://doi.org/10.1109/IEEM.2016.7798207.
[84] X. Li, Z. Tu, Q. Jia, X. Man, H. Wang, X. Zhang, Deep-level quality management
based on big data analytics with case study, Proc. - 2017 Chinese Autom. Congr.
CAC 2017. 2017-Janua, 2017, pp. 4921–4926, , https://doi.org/10.1109/CAC.
2017.8243651.
[85] M. Syafrudin, N.L. Fitriyani, D. Li, G. Alfian, J. Rhee, Y.S. Kang, An open source-
based real-time data processing architecture framework for manufacturing sus-
tainability, Sustain (2017) 9, https://doi.org/10.3390/su9112139.
[86] S. Venkatesh, Web-enabled real-time quality feedback for factory systems using
MTConnect, ASME 2012 Int. Des. Eng. Tech. Conf. Comput. Inf. Eng. Conf. 2012,
pp. 403–409.
[87] H.-M. Hou, J.-F. Kung, Y.-B. Hsu, Y. Yamazaki, K. Maruyama, Y. Toyoshima,
C. Chen, Prediction of ppm level electrical failure by using physical variation
analysis, Proc.SPIE. (2016) 9778, https://doi.org/10.1117/12.2229410.
[88] M.-K. Zheng, X.-G. Ming, X.-Y. Zhang, G.-M. Li, MapReduce based parallel baye-
sian network for manufacturing quality control, Chinese J. Mech. Eng. 30 (2017)
18
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
1216–1226, https://doi.org/10.1007/s10033-017-0179-0.
[89] S.A. Jacobs, A. Dagnino, Large-scale industrial alarm reduction and critical events
mining using graph analytics on spark, Proc. - 2016 IEEE 2nd Int. Conf. Big Data
Comput. Serv. Appl. BigDataService 2016, 2016, pp. 66–71, , https://doi.org/10.
1109/BigDataService.2016.21.
[90] H. Kemper, H. Baars, H. Lasi, An integrated business intelligence framework:
closing the gap between IT support for management and for production, Bus.
Intell. Perform. Manag. Springer, London, 2013, pp. 13–27, , https://doi.org/10.
1007/978-1-4471-4866-1.
[91] B.R. Ferrer, W.M. Mohammed, J.L.M. Lastra, A solution for processing supply
chain events within ontology-based descriptions, (2016) 4877–4883.
[92] M.J.A.G. Izaguirre, A. Lobov, J.L.M. Lastra, OPC-UA and DPWS interoperability
for factory floor monitoring using complex event processing, IEEE Int. Conf. Ind.
Informatics (2011) 205–210, https://doi.org/10.1109/INDIN.2011.6034874.
[93] J. Campos, P. Sharma, U.G. Gabiria, E. Jantunen, D. Baglee, A Big Data analytical
architecture for the asset management, Procedia CIRP 64 (2017) 369–374,
https://doi.org/10.1016/j.procir.2017.03.019.
[94] P. Gölzer, L. Simon, P. Cato, M. Amberg, Designing global manufacturing networks
using Big Data, Procedia CIRP 33 (2015) 191–196, https://doi.org/10.1016/j.
procir.2015.06.035.
[95] M. Zimmermann, U. Breitenbucher, M. Falkenthal, F. Leymann, K. Saatkamp,
Standards-based function shipping - How to use TOSCA for shipping and executing
data analytics software in remote manufacturing environments, Proc. - 2017 IEEE
21st Int. Enterp. Distrib. Object Comput. Conf. EDOC 2017, 2017, pp. 50–60, ,
https://doi.org/10.1109/EDOC.2017.16 2017-January.
[96] L.K. B, C. Gr, K. Jan, E. Hoos, C. Kiefer, C. Weber, S. Silcher, B. Mitschang, The
stuttgart IT architecture for manufacturing an architecture for the data-driven
factory, Int. Conf. Enterp. Inf. Syst. 2016, pp. 53–80, , https://doi.org/10.1007/
978-3-319-62386-3.
[97] W.M. Mohammed, B.R. Ferrer, L. Jose, M. Lastra, D. Aleixo, C. Agostinho,
Configuring and visualizing the data resources in a cloud-based data collection
framework, 2017 Int. Conf. Eng. Technol. Innov. Eng. Technol. Innov. Manag.
Beyond 2020 New Challenges, New Approaches, ICE/ITMC 2017 - Proc. 2018, pp.
1201–1208, , https://doi.org/10.1109/ICE.2017.8280017 2018-January.
[98] X. Ye, T.Y. Park, S.H. Hong, Y. Ding, A. Xu, Implementation of a production-
control system using integrated automation ML and OPC UA, 2018 Work. Metrol.
Ind. 4.0 IoT, MetroInd 4.0 IoT 2018 - Proc. 2018, pp. 242–247, , https://doi.org/
10.1109/METROI4.2018.8428310.
[99] Y. Cao, S. Wang, L. Kang, C. Li, L. Guo, Study on machining service modes and
resource selection strategies in cloud manufacturing, Int. J. Adv. Manuf. Technol.
81 (2015) 597–613, https://doi.org/10.1007/s00170-015-7222-z.
[100] S.K. Panda, T. Schroder, L. Wisniewski, C. Diedrich, PlugProduce integration of
components into OPC UA, based data-space, IEEE Int. Conf. Emerg. Technol. Fact.
Autom. ETFA, 2018, pp. 1095–1100, , https://doi.org/10.1109/ETFA.2018.
8502663 2018-September.
[101] F. Tao, L. Zhang, Y. Liu, Y. Cheng, L. Wang, X. Xu, Manufacturing service man-
agement in cloud manufacturing: overview and future research directions, J.
Manuf. Sci. Eng. 137 (2015) 040912, , https://doi.org/10.1115/1.4030510.
[102] H. Yang, M. Park, M. Cho, M. Song, S. Kim, A system architecture for manu-
facturing process analysis based on big data and process mining techniques, Big
Data (Big Data), 2014 IEEE Int. Conf. 2014, pp. 1024–1029, , https://doi.org/10.
1109/BigData.2014.7004336.
[103] J.C.C. Tseng, J.Y. Gu, P.F. Wang, C.Y. Chen, C.F. Li, V.S. Tseng, A scalable complex
event analytical system with incremental episode mining over data streams, 2016
IEEE Congr. Evol. Comput. CEC 2016, 2016, pp. 648–655, , https://doi.org/10.
1109/CEC.2016.7743854.
[104] B. Suryajaya, C.C. Chen, M.H. Hung, Y.Y. Liu, J.X. Liu, Y.C. Lin, A fast large-size
production data transformation scheme for supporting smart manufacturing in
semiconductor industry, IEEE Int. Conf. Autom. Sci. Eng. 2018, pp. 275–281, ,
https://doi.org/10.1109/COASE.2017.8256114 2017-Augus.
[105] A. Luckow, M. Cook, N. Ashcraft, E. Weill, E. Djerekarov, B. Vorster, Deep learning
in the automotive industry: Applications and tools, Big Data Int. Conf. Big Data.
2016, pp. 3759–3768, , https://doi.org/10.1109/BigData.2016.7841045.
[106] A. Brodsky, G. Shao, M. Krishnamoorthy, A. Narayanan, D. Menasce, R. Ak,
Analysis and optimization in smart manufacturing based on a reusable knowledge
base for process performance models, Proc. - 2015 IEEE Int. Conf. Big Data, IEEE
Big Data 2015, 2015, pp. 1418–1427, , https://doi.org/10.1109/BigData.2015.
7363902.
[107] S.C. Feng, W.Z. Bernstein, T. Hedberg, A. Barnard Feeney, Toward knowledge
management for smart manufacturing, J. Comput. Inf. Sci. Eng. 17 (2017) 031016,
, https://doi.org/10.1115/1.4037178.
[108] M. Stonebraker, I.F. Ilyas, Data integration: The current status and the way for-
ward, IEEE Data Eng. Bull. 41 (2018) 3–9.
[109] D. Zhang, B. Xu, J. Wood, Predict failures in production lines: A two-stage ap-
proach with clustering and supervised learning, Proc. - 2016 IEEE Int. Conf. Big
Data, Big Data 2016, 2016, pp. 2070–2074, , https://doi.org/10.1109/BigData.
2016.7840832.
[110] V. Jirkovský, M. Obitko, P. Novák, P. Kadera, Big data analysis for sensor time-
series in automation, 19th IEEE Int. Conf. Emerg. Technol. Fact. Autom. ETFA
2014, 2014, https://doi.org/10.1109/ETFA.2014.7005183.
[111] J. Bakakeu, J. Fuchs, T. Javied, M. Brossog, J. Franke, H. Klos, W. Eberlein,
S. Tolksdorf, J. Peschke, L. Jahn, Multi-Objective design space exploration for the
integration of advanced analytics in cyber-physical production systems, IEEE Int.
Conf. Ind. Eng. Eng. Manag. 2019, pp. 1866–1873, , https://doi.org/10.1109/
IEEM.2018.8607483 2019-December.
[112] R. Lynn, W. Louhichi, M. Parto, E. Wescoat, T. Kurfess, Rapidly deployable
Y. Cui, et al.
MTConnect-based machine tool monitoring systems, Proc. ASME 12TH Int. Manuf.
Sci. Eng. Conf. - 2017, 3 2017.
[113] D. Libes, S. Shin, J. Woo, Considerations and recommendations for data avail-
ability for data analytics for manufacturing, 3rd IEEE Int. Conf. Big Data, IEEE Big
Data, 2015, pp. 68–75, , https://doi.org/10.1109/BigData.2015.7363743.
[114] C. Zhao, L. Zhang, X.Z.L. Zhang, Cloud manufacturing resource management based
on metadata, ASME 2015 Int. Manuf. Sci. Eng. Conf. MSEC 2015, 2 2015, pp. 1–8,
, https://doi.org/10.1115/MSEC20159388.
[115] Y. Cheng, W. Shang, L. Zhu, D. Zhang, D. Feng, Items analysis of postal super-
vision, 2016 IEEE/ACIS 15th Int. Conf. Comput. Inf. Sci. ICIS 2016 - Proc. 2016,
pp. 3–5, , https://doi.org/10.1109/ICIS.2016.7550949.
[116] P. Gaj, K. Andrzej, P. Stera, Ontology-Based integrated monitoring of Hadoop
clusters in industrial environments with OPC UA and RESTful web services,
Commun. Comput. Inf. Sci. 522 (2015) 162–171, https://doi.org/10.1007/978-3-
319-19419-6.
[117] H. Derhamy, J. Ronnholm, J. Delsing, J. Eliasson, J. Van Deventer, Protocol in-
teroperability of OPC UA in service oriented architectures, Proc. - 2017 IEEE 15th
Int. Conf. Ind. Informatics, INDIN 2017, 2017, pp. 44–50, , https://doi.org/10.
1109/INDIN.2017.8104744.
[118] K.B. Lee, E.Y. Song, P.S. Gu, Integration of MTConnect and standard-based sensor
networks for manufacturing equipment monitoring, ASME 2012 Int. Manuf. Sci.
Eng. Conf. MSEC, 2012, pp. 4–8.
[119] A.N. Richter, T.M. Khoshgoftaar, S. Landset, T. Hasanin, A multi-dimensional
comparison of toolkits for machine learning with Big Data, Proc. - 2015 IEEE 16th
Int. Conf. Inf. Reuse Integr. IRI 2015, 2015, pp. 1–8, , https://doi.org/10.1109/
IRI.2015.12.
[120] A. Jos, Integration of sensors, controllers and instruments, Sensors (2017), https://
doi.org/10.3390/s17071512.
[121] A. Luckow, K. Kennedy, F. Manhardt, E. Djerekarov, B. Vorster, A. Apon,
Automotive big data: applications, workloads and infrastructures, Proc. - 2015
IEEE Int. Conf. Big Data, IEEE Big Data 2015, 2015, pp. 1201–1210, , https://doi.
org/10.1109/BigData.2015.7363874.
[122] C. Mathis, Data Lakes, Datenbank-Spektrum. (2017). doi:10.1007/s13222-017-
0272-7.
[123] M. Sarnovsky, P. Bednar, M. Smatana, Data integration in scalable data analytics
platform for process industries, 2017 IEEE 21st Int. Conf. Intell. Eng. Syst. 2017,
pp. 187–192, , https://doi.org/10.1109/INES.2017.8118553.
[124] M. Sarnovsky, P. Bednar, M. Smatana, Big data processing and analytics platform
architecture for process industry factories, Big Data Cogn. Comput. 2 (2018) 3,
https://doi.org/10.3390/bdcc2010003.
[125] G.L. Ooi, Y.-H. Wang, P.S. Tan, Z. Zhang, Y. Gao, J.K. Chow, Y. Wu, Q. Yuan,
Customizable and scalable geotechnical laboratory testing and field monitoring
with new sensing and big data technologies, ICSMGE 2017 - 19th Int. Conf. Soil
Mech. Geotech. Eng. 2017, pp. 471–474.
[126] H. Liang, L. Feng, Z. Chun, Application of the Big Data technology for massive data
of the whole life cycle of EMU, in: F. Xhafa, S. Patnaik, Z. Yu (Eds.), Recent Dev.
Intell. Syst. Interact. Appl. Springer International Publishing, Cham, 2017, pp.
219–224.
[127] L. Zheng, L. Tang, T. Li, B. Duan, M. Lei, P. Wang, C. Zeng, L. Li, Y. Jiang, W. Xue,
J. Li, C. Shen, W. Zhou, H. Li, Applying data mining techniques to address critical
process optimization needs in advanced manufacturing, Proc. 20th ACM SIGKDD
Int. Conf. Knowl. Discov. Data Min. - KDD ’14, 2014, pp. 1739–1748, , https://doi.
org/10.1145/2623330.2623347.
[128] S. Windmann, A. Maier, O. Niggemann, C. Frey, A. Bernardi, Y. Gu, H. Pfrommer,
T. Steckel, M. Krüger, R. Kraus, Big data analysis of manufacturing processes, J.
Phys. Conf. Ser. (2015) 659, https://doi.org/10.1088/1742-6596/659/1/012055.
[129] R. Bohlin, L. Lindkvist, J. Hagmar, J.S. Carlson, K. Bengtsson, Data flow and
communication framework supporting digital twin for geometry assurance Robert,
Proc. ASME 2017 Int. Mech. Eng. Congr. Expo. 2017, pp. 1–7.
[130] K.E. Harper, J. Zheng, S.A. Jacobs, A. Dagnino, A. Jansen, T. Goldschmidt,
A. Marinakis, Industrial analytics pipelines, Proc. - 2015 IEEE 1st Int. Conf. Big
Data Comput. Serv. Appl. BigDataService 2015, 2015, pp. 242–248, , https://doi.
org/10.1109/BigDataService.2015.38.
[131] T. Goldschmidt, M.K. Murugaiah, C. Sonntag, B. Schlich, S. Biallas, P. Weber,
Cloud-based control : a multi-tenant, horizontally scalable Soft-PLC, 2015 IEEE 8th
Int. Conf. Cloud Comput. 2015, https://doi.org/10.1109/CLOUD.2015.124.
[132] H.-S. Park, J.-H. Kim, C.-H. Choi, B.-R. Jung, K.-H. Lee, S.-Y. Chi, W.-S. Cho, In-
Memory data grid system for real-time processing of machine sensor data in a
smart factory environment, Proc. 2015 Int. Conf. Big Data Appl. Serv. - BigDAS
’15, 2015, pp. 92–97, , https://doi.org/10.1145/2837060.2837073.
[133] N. Ramakrishnanus, R. Ghosh, Distributed dynamic elastic nets : a scalable ap-
proach for regularization in dynamic manufacturing environments, 2015 IEEE Int.
Conf. Big Data, 2015, pp. 2752–2761.
[134] S. Wang, J. Ouyang, D. Li, C. Liu, An integrated industrial ethernet solution for the
implementation of smart factory, IEEE Access (2017) 25455–25462, https://doi.
org/10.1109/ACCESS.2017.2770180.
[135] S. Division, C. Hsing, N. Village, N. City, N. County, M. Availability, Developing a
cloud virtual maintenance system for machine tools management, 2015 11th Int.
Conf. Heterog. Netw. Qual. Reliab. Secur. Robustness. 2015, pp. 358–364.
[136] J. Wan, S. Tang, D. Li, S. Wang, C. Liu, H. Abbas, S. Member, A.V. Vasilakos, A
manufacturing big data solution for active preventive maintenance, IEEE Trans.
Ind. INFORMATICS, 2017, pp. 2039–2047.
[137] J.-H. Ku, A study on prediction model of equipment failure through analysis of Big
Data based on RHadoop, Wirel. Pers. Commun. 98 (2017) 3163–3176, https://doi.
org/10.1007/s11277-017-4151-1.
[138] W. Lee, J. Cho, L. Lee, S. Korea, Time series abnormal data detection for smart
19
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
factory, Int. J. Control Autom. 11 (2018) 91–98.
[139] H. Chen, X. Fei, S. Wang, X. Lu, G. Jin, W. Li, X. Wu, Energy consumption data
based machine anomaly detection, 2014 Second Int. Conf. Adv. Cloud Big Data,
2014, pp. 136–142, , https://doi.org/10.1109/CBD.2014.24.
[140] N. Stojanovic, M. Dinic, L. Stojanovic, A data-driven approach for multivariate
contextualized anomaly detection: industry use case, Proc. - 2017 IEEE Int. Conf.
Big Data, Big Data 2017. 2018-Janua, 2018, pp. 1560–1569, , https://doi.org/10.
1109/BigData.2017.8258090.
[141] M. Canizo, E. Onieva, A. Conde, S. Charramendieta, S. Trujillo, Real-time pre-
dictive maintenance for wind turbines using Big Data frameworks, IEEE Int. Conf.
Progn. Heal. Manag. 2017, pp. 1–8, , https://doi.org/10.1109/ICPHM.2017.
7998308.
[142] T. Zaarour, N. Pavlopoulou, S. Hasan, U. ul Hassan, E. Curry, Grand challenge:
automatic anomaly detection over sliding windows, Proc. 11th ACM Int. Conf.
Distrib. Event-Based Syst. - DEBS ’17, 2017, pp. 310–314, , https://doi.org/10.
1145/3093742.3095105.
[143] J.H. Lee, S. Do Noh, H.-J. Kim, Y.-S. Kang, Implementation of cyber-physical
production systems for quality prediction and operation control in metal casting,
Sensors (Switzerland) (2018) 18, https://doi.org/10.3390/s18051428.
[144] A. Stojadinović, Industry Paper : Dynamic monitoring for improving worker safety
at the workplace: uUse case from a manufacturing shop floor, (2015) 205–216.
[145] H. Haskamp, F. Orth, J. Wermann, A.W. Colombo, Implementing an OPC UA in-
terface for legacy PLC-based automation systems using the Azure cloud: An ICPS-
architecture with a retrofitted RFID system, Proc. - 2018 IEEE Ind. Cyber-Physical
Syst. ICPS 2018, 2018, pp. 115–121, , https://doi.org/10.1109/ICPHYS.2018.
8387646.
[146] D. Cemernek, H. Gursch, R. Kern, Big data as a promoter of industry 4.0: Lessons of
the semiconductor industry, Proc. - 2017 IEEE 15th Int. Conf. Ind. Informatics,
INDIN 2017, 2017, pp. 239–244, , https://doi.org/10.1109/INDIN.2017.8104778.
[147] C. Ellwein, O. Riedel, O. Meyer, D. Schel, Rent'n’Produce: A secure cloud manu-
facturing platform for small and medium enterprises, 2018 IEEE Int. Conf. Eng.
Technol. Innov. ICE/ITMC 2018 - Proc. 2018, pp. 1–6, , https://doi.org/10.1109/
ICE.2018.8436332.
[148] N. Ferry, G. Terrazas, P. Kalweit, A. Solberg, S. Ratchev, D. Weinelt, Towards a big
data platform for managing machine generated data in the cloud, Proc. - 2017
IEEE 15th Int. Conf. Ind. Informatics, INDIN 2017, 2017, pp. 263–270, , https://
doi.org/10.1109/INDIN.2017.8104782.
[149] A. Angrish, B. Starly, Y.S. Lee, P.H. Cohen, A flexible data schema and system
architecture for the virtualization of manufacturing machines (VMM), J. Manuf.
Syst. 45 (2017) 236–247, https://doi.org/10.1016/j.jmsy.2017.10.003.
[150] R.S. Peres, A.D. Rocha, A. Coelho, J. Barata Oliveira, A. Highly Flexible,
Distributed data analysis framework for industry 4.0 manufacturing systems, in:
T Borangiu, D. Trentesaux, A. Thomas, P. Leitão, J.B. Oliveira (Eds.), Serv.
Orientat. Holonic Multi-Agent Manuf. Springer International Publishing, Cham,
2017, pp. 373–381.
[151] M.Y. Santos, J.O. e Sá, C. Costa, J. Galvão, C. Andrade, B. Martinho, F.V. Lima,
E. Costa, A big data analytics architecture for industry 4.0, WorldCIST 2017
Recent Adv. Inf. Syst. Technol, 2017, https://doi.org/10.1007/978-3-319-56538-
5 0.
[152] P. Tanuska, L. Spendla, M. Kebisek, Data integration for incidents analysis in
manufacturing infrastructure, Comput. Conf. 2017, 2017, pp. 340–345.
[153] H. Lin, J.A. Harding, C. Chen, A hyperconnected manufacturing collaboration
system using the semantic web and Hadoop ecosystem system, Procedia CIRP 52
(2016) 18–23, https://doi.org/10.1016/j.procir.2016.07.075.
[154] C. Gröger, Building an industry 4.0 analytics platform, Datenbank-Spektrum.
(2018). doi:10.1007/s13222-018-0273-1.
[155] M.C. Domenech, L.P. Rauta, M.D. Lopes, P.H. Da Silva, R.C. Da Silva, B.W. Mezger,
M.S. Wangham, Providing a smart industrial environment with the web of things
and cloud computing, Proc. - 2016 IEEE Int. Conf. Serv. Comput. SCC 2016, 2016,
pp. 641–648, , https://doi.org/10.1109/SCC.2016.89.
[156] J. Wan, S.,. Tang, D. Li, S. Wang, C. Liu, H. Abbas, A.V. Vasilakos, A manufacturing
Big Data solution for active preventive maintenance, IEEE Trans. Ind. Informatics.
13 2017, pp. 2039–2047, , https://doi.org/10.1201/b15906-13.
[157] N. Rivetti, Y. Busnel, A. Gal, Grand challenge: Flinkman - Anomaly detection in
manufacturing equipment with Apache Flink, Proc. 11th ACM Int. Conf. Distrib.
Event-Based Syst. - DEBS ’17, 2017, pp. 274–279, , https://doi.org/10.1145/
3093742.3095099.
[158] M. Canizo, E. Onieva, A. Conde, S. Charramendieta, S. Trujillo, Real-time pre-
dictive maintenance for wind turbines using Big Data frameworks, (2017) 1–8.
[159] M.R. Brule, Big data in EP: Real-time adaptive analytics and data-flow archi-
tecture, Soc. Pet. Eng. - SPE Digit. Energy Conf. Exhib. 2013, 2013, pp. 305–311, ,
https://doi.org/10.2118/163721-MS.
[160] D. Wu, J. Terpenny, L. Zhang, R. Gao, T. Kurfess, Fog-enabled architecture for
data-driven cyber-manufacturing systems, ASME 2016 11th Int. Manuf. Sci. Eng.
Conf. MSEC 2016, 2 2016, https://doi.org/10.1115/MSEC2016-8559
V002T04A032.
[161] R. Gao, L. Wang, R. Teti, D. Dornfeld, S. Kumara, M. Mori, M. Helu, Cloud-enabled
prognosis for manufacturing, CIRP Ann. - Manuf. Technol. 64 (2015) 749–772,
https://doi.org/10.1016/j.cirp.2015.05.011.
[162] Y. Wu, S. Wang, Streaming analytics processing in manufacturing performance
monitoring and prediction, 2017 IEEE Int. Conf. Big Data (Big Data) (2017)
3285–3289.
[163] Y. Zhang, S. Ren, Y. Liu, T. Sakao, D. Huisingh, A framework for Big Data driven
product lifecycle management, J. Clean. Prod. 159 (2017) 229–240, https://doi.
org/10.1016/j.jclepro.2017.04.172.
[164] M. Naeem, N. Moalla, Y. Ouzrout, A. Bouaras, An ontology based digital
Y. Cui, et al.
preservation system for enterprise collaboration, 2014 IEEE/ACS 11th Int. Conf.
Comput. Syst. Appl. 2014, pp. 691–698, , https://doi.org/10.1109/AICCSA.2014.
7073267.
[165] Q.P. He, J. Wang, D. Shah, N. Vahdat, Statistical process monitoring for IoT-en-
abled cybermanufacturing: Opportunities and challenges, IFAC-PapersOnLine 50
(2017) 14946–14951, https://doi.org/10.1016/j.ifacol.2017.08.2546.
[166] F. Tao, J. Cheng, Q. Qi, IIHub: An industrial internet-of-things hub toward smart
manufacturing based on cyber-physical system, IEEE Trans. Ind. Informatics. 14
2018, pp. 2271–2280, , https://doi.org/10.1109/TII.2017.2759178.
[167] S. Ramírez-Gallego, A. Fernández, S. García, M. Chen, F. Herrera, Big Data,
Tutorial and guidelines on information and process fusion for analytics algorithms
with MapReduce, Inf. Fusion. 42 (2018) 51–61, https://doi.org/10.1016/j.inffus.
2017.10.001.
[168] S. Michael, SQL databases v. NoSQL databases, Commun. ACM. 53 (2010) 10–11,
https://doi.org/10.1145/1721654.1721659.
[169] I. Kovačević, I. Mekterović, Alternative business intelligence engines, Inf.
Commun. Technol. Electron. Microelectron. 2017, pp. 1617–1622, , https://doi.
org/10.23919/MIPRO.2017.7973638.
[170] P. Kannan, Beyond Hadoop Mapreduce Apache Tez and Apache Spark, (2015).
[171] S. Chintapalli, D. Dagit, B. Evans, R. Farivar, T. Graves, M. Holderbaugh, Z. Liu,
K. Nusbaum, K. Patil, B.J. Peng, P. Poulosky, Benchmarking streaming computa-
tion engines: Storm, flink and spark streaming, Proc. - 2016 IEEE 30th Int. Parallel
Distrib. Process. Symp. IPDPS 2016, 2016, pp. 1789–1792, , https://doi.org/10.
1109/IPDPSW.2016.138.
[172] D. Abadi, R. Agrawal, A. Ailamaki, M. Balazinska, P.A. Bernstein, M.J. Carey,
S. Chaudhuri, J. Dean, A. Doan, M.J. Franklin, J. Gehrke, The Beckman report on
database research, Commun. ACM. 59 (2016) 92–99.
[173] S.V. Ranawade, Online analytical processing on hadoop Hadoop using Apache
Kylin, 12 (2017) 1–5.
[174] M. Islam, A.K. Huang, M. Battisha, M. Chiang, S. Srinivasan, C. Peters,
A. Neumann, A. Abdelnur, Oozie: towards a scalable workflow management
system for Hadoop, Proc. 1st ACM SIGMOD Work. Scalable Work. Exec. Engines
Technol. - SWEET ’12, 2012, pp. 1–10, , https://doi.org/10.1145/2443416.
2443420.
[175] K. Sundaravarathan, P. Martin, D. Rope, M. McRoberts, C. Statchuk, MEWSE:
multi-engine workflow submission and execution on Apache YARN, Proc. 26th
Annu. Int. Conf. Comput. Sci. Softw. Eng. 2016, pp. 194–200 http://dl.acm.org/
citation.cfm?id=3049877.3049897.
[176] I. Suriarachchi, B. Plale, Crossing analytics systems : a case for integrated prove-
nance in data lakes, 2016 IEEE 12th Int. Conf. e-Science Crossing, Baltimore, MD,
USA, IEEE, 2016, pp. 349–354, , https://doi.org/10.1109/eScience.2016.
7870919.
[177] M.O. Gökalp, K. Kayabay, M. Zaki, A. Koçyiğit, P.E. Eren, A. Neely, Big-Data data
analytics architecture for businesses : a comprehensive review on new open-source
big-data tools, (2017).
[178] B. Stein, A. Morrison, The enterprise data lake: Better integration and deeper
analytics, PWC Technol. Forecast Rethink. Integr. 2014 http://www.pwc.com/us/
en/technology-forecast/2014/cloud-computing/assets/pdf/pwc-technology-
forecast-data-lakes.pdf.
[179] N. Miloslavskaya, A. Tolstoy, Big Data, fast data and data lake concepts, Procedia
Comput. Sci. 88 (2016) 300–305, https://doi.org/10.1016/j.procs.2016.07.439.
[180] R. Bose, V. Sugumaran, Application of knowledge management technology in
customer relationship management, 10 (2003) 3–17. doi:10.1002/kpm.163.
[181] B.R. Ferrer, W.M. Mohammed, J.L.M. Lastra, A solution for processing supply
chain events within ontology-based descriptions, IECON Proc. (Industrial Electron.
Conf. 2016, pp. 4877–4883, , https://doi.org/10.1109/IECON.2016.7793020.
[182] M. Mercier, D. Glesser, Y. Georgiou, O. Richard, Big Data and HPC collocation :
using HPC idle resources for Big Data analytics, Big Data (Big Data), 2017 IEEE Int.
Conf. 2017, pp. 347–352.
[183] G. Adamson, L. Wang, M. Holm, P. Moore, Cloud manufacturing – a critical review
of recent development and future trends, Int. J. Comput. Integr. Manuf. (2017)
1–34, https://doi.org/10.1080/0951192X.2015.1031704.
[184] J. Lee, H.D. Ardakani, S. Yang, B. Bagheri, Industrial Big Data analytics and cyber-
physical systems for future maintenance & service innovation, Procedia CIRP 38
(2015) 3–7, https://doi.org/10.1016/j.procir.2015.08.026.
[185] E. Knill, Quantum computing with realistically noisy devices, Nature 434 (2005)
39–44, https://doi.org/10.1038/nature03350.
[186] Deutsche kommission elektrotechnik DKE; din DIN e.V., german German
20
Robotics and Computer Integrated Manufacturing 62 (2020) 101861
standardization roadmap industry 4.0, (2016) 523.
[187] W. and S. Kagermann, H., Riemensperger, F., Hoke, D., Helbig, J., Stocksmeier, D.,
Wahlster, Recommendations for the strategic initiative coordination and editing,
2015.
[188] NIST, Framework for improving critical infrastructure cybersecurity, 2018. doi:10.
1109/JPROC.2011.2165269.
[189] P. Wang, R.X. Gao, Z. Fan, Cloud computing for cloud manufacturing: Benefits and
limitations, J. Manuf. Sci. Eng. 137 (2015) 044002, , https://doi.org/10.1115/1.
4030209.
[190] H. Fang, Managing data lakes in Big Data era, Cyber Technol. Autom. Control.
Intell. Syst. (CYBER), 2015 IEEE Int. Conf. 2015, pp. 820–824.
[191] D. Bernstein, Containers and cloud: From LXC to docker to kubernetes, IEEE Cloud
Comput 1 (2014) 81–84, https://doi.org/10.1109/MCC.2014.51.
[192] K. Stouffer, J. Falco, K. Kent, Guide to supervisory control and data acquisition
(SCADA) and industrial control systems security, NIST Spec. Publ. SP800-82,
2006, pp. 800–882, , https://doi.org/10.6028/NIST.SP.800.82.
[193] S. Nazir, S. Patel, D. Patel, Assessing and augmenting SCADA cyber security: A
survey of techniques, Comput. Secur. 70 (2017) 436–454, https://doi.org/10.
1016/j.cose.2017.06.010.
[194] H. Shih, W, Ludwig, The biggest challenges of data-driven manufacturing, Harv.
Bus. Rev. (2016), https://hbr.org/2016/05/the-biggest-challenges-of-data-driven-
manufacturing accessed June 6, 2019.
[195] Allae Erraissi, Abdessamad Belangour, Abderrahim Tragha, Digging into Hadoop-
based Big Data architectures, Int. J. Comput. Sci. Issues IJCSI 14 (2017) 52–59,
https://doi.org/10.20943/01201706.5259.
[196] H. Fleischmann, S. Spreng, J. Kohl, D. Kisskalt, J. Franke, Distributed condition
monitoring systems in electric drives manufacturing, 2016 6th Int. Electr. Drives
Prod. Conf. EDPC 2016 - Proc. 2016, pp. 52–57, , https://doi.org/10.1109/EDPC.
2016.7851314.
[197] J. Manyika, M. Chui, J. Bughin, R. Dobbs, P. Bisson, Marrs, Disruptive technolo-
gies: Advances that Will Transform Life, Business, and the Global Economy,
McKinsey Glob. Insitute, 2013, p. 163 http://www.mckinsey.com/insights/
business_technology/disruptive_technologies%5Cnhttp://www.chrysalixevc.com/
pdfs/mckinsey_may2013.pdf.
[198] A. Burger, H. Koziolek, J. Rückert, M. Platenius-Mohr, G. Stomberg, Bottleneck
identification and performance modeling of OPC UA opc ua communication
models, (2019) 231–242. doi:10.1145/3297663.3309670.
[199] Deutsche kommission elektrotechnik DKE; din DIN e.V., german German stan-
dardization roadmap industry 4.0, 2016.
[200] D. Loshin, Data integration, Bus. Intell. (2013) 189–210, https://doi.org/10.1016/
B978-0-12-385889-4.00013-2.
[201] J.M. Gutierrez-Guerrero, J.A. Holgado-Terriza, IMMAS an industrial meta-model
for automation system using OPC UA, Elektron, Ir Elektrotechnika 23 (2017) 3–11,
https://doi.org/10.5755/j01.eie.23.3.18324.
[202] Y.H. Wu, S. De Wang, L.J. Chen, C.J. Yu, Streaming analytics processing in
manufacturing performance monitoring and prediction, Proc. - 2017 IEEE Int.
Conf. Big Data, Big Data 2017, 2018, pp. 3285–3288, , https://doi.org/10.1109/
BigData.2017.8258312 2018-January.
[203] A. Stojadinović, N. Stojanović, L. Stojanović, Dynamic monitoring for improving
worker safety at the workplace, Proc. 9th ACM Int. Conf. Distrib. Event-Based Syst.
- DEBS ’15, 2015, pp. 205–216, , https://doi.org/10.1145/2675743.2771881.
[204] S. Scholze, K. Nagorny, R. Siafaka, K. Krone, An approach for cloud-based situa-
tional analysis for factories providing real-time reconfiguration services, in: 2017:
pp. 118–127. doi:10.1007/978-3-319-65151-4.
[205] L. Angrisani, G. Ianniello, I. Elettrica, N. Federico, Cloud based system for mea-
surement data management in large scale electronic production, Euro Med Telco
Conference (EMTC), IEEE, 2014, pp. 1–4.
[206] G. Hesse, B. Reissaus, C. Matthies, M. Lorenz, M. Kraus, M. Uflacker, Senska –
Towards an enterprise streaming benchmark, 2018. doi:10.1007/978-3-319-
72401-0_3.
[207] L. Banica, A. Hagiu, 4 - Using big data analytics to improve decision-making in
apparel supply chains A2 - Choi, Tsan-Ming BT - information systems for the
fashion and apparel industry, Woodhead Publ. Ser. Text. Woodhead Publishing,
2016, pp. 63–95 https://doi.org/10.1016/B978-0-08-100571-2.00004-X.
[208] S. Saeidlou, M. Saadat, E.A. Sharifi, D. Guiovanni, An ontology-based intelligent
data query system in manufacturing networks, Prod. Manuf. Res. 3277 (2017)
1–18, https://doi.org/10.1080/21693277.2017.1374887.